MI IVE API

REVISION HISTORY¶

Revision No.	Description	Date
2.03	Initial release.	04/12/2018
2.04	Added function descriptions in regard to MI_IVE_Bernsen, MI_IVE_LineFilterHor, MI_IVE_LineFilterVer, MI_IVE_NoiseRemoveHor, MI_IVE_NoiseRemoveVer, MI_IVE_AdpThresh, MI_IVE_Resize, MI_IVE_BAT, and MI_IVE_Acc.	06/05/2019
2.05	Added function descriptions to MI_IVE_Matrix_Transform and MI_IVE_Image_Dot. Added description of new input parameter for add function. Added description of new input mode for lbp function.	10/09/2019
2.06	Added function description to MI_IVE_Shift_Detector.	09/16/2020
2.07	Modified the description of MI_IVE_Csc. Modified the parameters range for matrix transform function. Modified the description of return value. Added API reference: MI_IVE_Shift_detector; MI_IVE_Alphablending.	10/14/2020
2.08	Correct the formula of MagAndAng and CannyHsysEdge. Add description of MI_IVE_MatrTranfCtrlMode_e.	08/13/2021
2.09	More detailed description of MI_IVE_Ccl calling method. Added fourth paragraph topic---Hardware acceleration implementation list.	08/26/2021
2.10	Add note for matrix_transform.	01/26/2022
2.11	Add new support enMode for CSC.	03/15/2022
2.12	Added function description to MI_IVE_Create_Handle. Rename IVE lib.	03/18/2022
2.13	Fix format error.	04/22/2022
2.14	Add description about limitation.	06/02/2022
2.15	Add new API reference of MI_IVE_BGBlur. Modify description of MI_IVE_Ccl & MI_IVE_Gmm & MI_IVE_Resize. Add new datatype of MI_IVE_HANDLE_MAX & MI_IVE_ResizeMethod_e & MI_IVE_BgBlurMode_e & MI_IVE_BgBlurCtrl_t.	12/13/2022
2.16	Update the maximum resolution limitation of MI_IVE_Resize.	03/30/2023
2.17	Add Mask preprocess mode of BGBlur.	07/31/2023
2.18	Revise and refine formulas in document. Refine the figure of coefficient template & MI_IVE_Thresh & MI_IVE_Gmm, and add the figure of MI_IVE_ThreshS16 & MI_IVE_ThreshU16. Move the common part in MI_IVE_Filter to the front of API reference. Modify the description of MI_IVE_Ccl & MI_IVE_Gmm & MI_IVE_Resize & MI_IVE_BGBlur.	08/22/2023
2.19	Refine the description and add the example for MI_IVE_Filter & MI_IVE_OrdStatFiler & MI_IVE_Resize. Add new section 5.PROCFS INTRODUCTION.	04/23/2024
2.20	Document optimization.	04/18/2025

1. OVERVIEW¶

1.1. Module Description¶

MI_IVE（Intelligent Video Engine）refers to the image/video engine module, mainly providing the API of image algorithms. It can be used for tasks such as encoding, decoding, feature extraction, object detection, face recognition, image enhancement, image denoising, and motion detection of images and videos. There are two implementation sources. One is HW IVE, which is a hardware implementation and is used by calling the driver. One is SW IVE, which is implemented by software and used through the mve interface. The priority of the execution operators is: HW IVE > Neon > PureC.

1.2. Basic Structure¶

MI_IVE hardlinks the key software operators or common operators that consume the most CPU resources and are used in image and intelligent video analysis to form an image coprocessor. Before hardening, the operators are calculated using SW ALG. After hardening, the operators are calculated using HW. Therefore, there will be some changes in some resolution sizes, parameter ranges and the required memory space size. Compared with the previous calculation using SW ALG, the calculation using HW can improve the performance of the operator, thereby enhancing the ability and computing speed of the entire chip for video analysis.

Figure 1‑1 MI_IVE Module Architecture

1.3. Function Introduction¶

MI_IVE supports the following functions:

Support the execution of HW/SW operators
Support the switching of hw<->sw operators
Supports multiple input/output image resolutions and picture formats

1.4. Application Scenarios¶

MI_IVE can be applied to the following scenarios：

Linux scenarios

In the Linux environment, when users/modules call the MI IVE API, they need to link libmi_ive.a/so. When using non-public MI IVE APIs, they need to include "mi_ive_internal.h" to call the corresponding API.
Rtos scenarios

In the rtos environment, mi_ive.lib will be placed in the rtos binary. During the execution of the API, based on the selected defconfig configuration, it will be confirmed by MI_SYSCALL_SOCID to be executed into the code of mi_ive.ko or mi_ive.lib.

1.5. Chip Difference¶

The current is the description of the pcupid series chip.

Table 1-1 Implementation table of Intelligent Acceleration Hardware API

API Name	Ispahan	Pudding/Tiramisu/Muffin	Souffle	iFord	pcupid
MI_IVE_Csc		v	v	v
MI_IVE_FilterAndCsc		v	v	v
MI_IVE_Filter		v	v	v	v
MI_IVE_MagAndAng		v	v	v
MI_IVE_Hist		v	v	v
MI_IVE_Integ		v	v	v
MI_IVE_Dilate		v	v	v	v
MI_IVE_Erode		v	v	v	v
MI_IVE_Map		v	v	v
MI_IVE_And	v	v	v	v
MI_IVE_Or	v	v	v	v
MI_IVE_Xor	v	v	v	v
MI_IVE_Add	v	v	v	v
MI_IVE_Sub	v	v	v	v
MI_IVE_Ncc		v	v	v
MI_IVE_Sobel		v	v	v
MI_IVE_Thresh	v	v	v	v
MI_IVE_ThreshS16	v	v	v	v
MI_IVE_ThreshU16	v	v	v	v
MI_IVE_16BitTo8Bit		v	v	v
MI_IVE_OrdStatFilter		v	v	v	v
MI_IVE_Sad	v	v	v	v
MI_IVE_Bernsen		v	v	v	v
MI_IVE_Lbp		v	v	v
MI_IVE_BAT		v	v	v
MI_IVE_AdpThresh		v	v	v
MI_IVE_Matrix_Transform		v	v	v
MI_IVE_Image_Dot		v	v	v
MI_IVE_AlphaBlending		v	v	v
MI_IVE_Ccl			v
MI_IVE_Gmm			v
MI_IVE_Resize			v		v
MI_IVE_BGBlur			v	v

1.6. Working Principle¶

1.6.1 API Process¶

Figure 1‑1 MI_IVE API Process

1.6.2 Description of the intelligent acceleration engine flush and Invalidate mechanism¶

The intelligent acceleration engine supports pure software, Neon acceleration and hardware acceleration. When continuously calling the IVE API, it internally supports automatic Cache Invalidate and Cache Flush of the MI_IVE_Image_t structure variable. Only before and after calling the hardware acceleration operator and before and after the CPU's intervention in the operation, Cache Invalidate or Cache Flush needs to be performed independently. The following example illustrates that a MI_IVE_Image_t src is declared; src.u16Reserved is initialized to 0. During the operation process of the intelligent acceleration engine, this flag is used to record the status of software and hardware. If the IVE API is continuously called (as shown on the left in the figure), even if MI_IVE_Ccl is purely soft operation, It still supports automatic Cache Invalidate and Cache Flush internally. If the memory is intervened by the CPU for operation (as shown on the right in the figure), or when the API call ends, the user must perform a Cache Invalidate or Cache Flush by themselves and set src.u16Reserved to 0. The hardware acceleration supported by different chips should refer to Table 1-1.

1.6.3 Instructions on Usage Restrictions of the Intelligent Acceleration Engine¶

address : 16-byte alignment
stride : 16-pixel alignment
width : 16-pixel alignment，MI_IVE_BGBlursupport 2-pixel alignment
height : 2-pixel alignment

1.6.4 Confirmation Instructions for the use of the Intelligent acceleration Engine¶

Can check the register of IVE for confirmation. Each function has its own corresponding code.

Table 1-2 IVE Register Info-1

	Ispahan	Pudding/Tiramisu/Muffin	Souffle/iFord/pcupid
BANK	1521	1191	151F
OFFSET	0004	0004	0004

Table 1-3 IVE Register Info-2

op_type	Function
0x00	MI_IVE_Filter
0x01	MI_IVE_Csc
0x02	MI_IVE_FilterAndCsc
0x03	MI_IVE_Sobel
0x04	MI_IVE_MagAndAng
0x05	MI_IVE_OrdStatFilter
0x06	MI_IVE_Bernsen
0x07	MI_IVE_Dilate
0x08	MI_IVE_Erode
0x09	MI_IVE_Thresh
0x0A	MI_IVE_ThreshS16
0x0B	MI_IVE_ThreshU16
0x0C	MI_IVE_And
0x0D	MI_IVE_Or
0x0E	MI_IVE_Xor
0x0F	MI_IVE_Add
0x10	MI_IVE_Sub
0x11	MI_IVE_16BitTo8Bit
0x12	MI_IVE_Map
0x13	MI_IVE_Hist
0x14	MI_IVE_Integ
0x15	MI_IVE_Sad
0x16	MI_IVE_Ncc
0x18	MI_IVE_Lbp
0x1F	MI_IVE_Matrix_Transform
0x20	MI_IVE_Image_Dot
0x21	MI_IVE_AlphaBlending
0x22	MI_IVE_Ccl
0x23	MI_IVE_Gmm
0x24	MI_IVE_Resize
0xFF	MI_IVE_BGBlur

1.7. Development Process¶

1.7.1 Compilation Configuration¶

Enter the root directory of alkaid project，make menuconfig
Press Enter to enter the Sdk Config sub-options
Press Enter to enter the Interface Compile Config sub-options
Press the space bar to select the ive submodule，and recompile project

After the compilation is completed, mi_ive.ko will be generated under sdk/interface/src/ive. Meanwhile, mi_ive.h and mi_ive_datatype.h will be released to the project/release directory. In the pure linux environment, it will be packaged into images by default. In the dualos environment, it is not packaged by default and needs to be manually compiled and installed.

1.8. Example Introduction¶

Take the use of the Filter operator as an example: MI_IVE_Filter

2. API REFERENCE¶

The MI IVE module provides the following APIs:

Name of API	Function
MI_IVE_Create_Handle	Create IVE handle index
MI_IVE_Create	Create an IVE handle
MI_IVE_Destroy	Destroy an IVE handle
MI_IVE_Filter	Execute a 5x5 template filter task
MI_IVE_Csc	Execute a color space conversion task
MI_IVE_FilterAndCsc	Execute a composite task of template filter plus color space conversion
MI_IVE_Sobel	Execute a 5x5 template Sobel-like gradient calculation task
MI_IVE_MagAndAng	Execute a 5x5 template gradient magnitude and angle calculation task
MI_IVE_Dilate	Execute a Dilate task
MI_IVE_Erode	Execute an Erode task
MI_IVE_Thresh	Execute an image thresholding task
MI_IVE_And	Execute an And task
MI_IVE_Sub	Execute a Subtract task
MI_IVE_Or	Execute an Or task
MI_IVE_Integ	Execute an integral graph statistics task
MI_IVE_Hist	Execute a histogram statistics task
MI_IVE_ThreshS16	Execute an S16 data to 8-bit data thresholding task
MI_IVE_ThreshU16	Execute a U16 data to U8 data thresholding task
MI_IVE_16BitTo8Bit	Execute a 16-bit data to 8-bit data linear transformation task
MI_IVE_OrdStatFilter	Execute a 3x3 template sequential statistic filtering task
MI_IVE_Map	Execute a Map (U8->U8 mapping assignment) task
MI_IVE_EqualizeHist	Execute a grayscale-image histogram equalization calculation task
MI_IVE_Add	Execute a weighted addition calculation task against two grayscale images
MI_IVE_Xor	Execute an XOR calculation task against two binary graphs
MI_IVE_Ncc	Execute a normalized cross-correlation calculation task against two images of the same resolution
MI_IVE_Ccl	Execute a connected region label task against binary images
MI_IVE_Gmm	Execute a GMM background modeling task
MI_IVE_CannyHysEdge	Execute a Canny strong edge extraction task against grayscale images
MI_IVE_CannyEdge	Execute the second half of Canny strong edge extraction task: connecting edge points to form a Canny edge map
MI_IVE_Lbp	Execute an LBP calculation task
MI_IVE_NormGrad	Execute a normalized gradient calculation task, in which the gradient average components are normalized to S8
MI_IVE_LkOpticalFlow	Execute a single layer LK optical flow calculation task
MI_IVE_Sad	Calculate 16-bit/8-bit SAD images of 4x4/8x8/16x16 blocks for two images, and threshold output for SAD
MI_IVE_Bernsen	Execute a Bernsen thresholding task for the 3x3 and 5x5 windows.
MI_IVE_LineFilterHor	Execute a horizontal density filter task for binary images.
MI_IVE_LineFilterVer	Execute a vertical density filter task for binary images.
MI_IVE_NoiseRemoveHor	Execute a horizontal noise removal task for binary images.
MI_IVE_NoiseRemoveVer	Execute a vertical noise removal task for binary images.
MI_IVE_AdpThresh	Execute an adaptive thresholding task.
MI_IVE_Resize	Execute an image scaling task.
MI_IVE_BAT	Execute the horizontal or vertical alternating time for binary images.
MI_IVE_Acc	Execute an accumulation task for two gray-scale images.
MI_IVE_Matrix_Transform	Execute the operation of matrix multiplication.
MI_IVE_Image_Dot	Execute the operation of dot product.
MI_IVE_Shift_Detector	Execute the operation of object tracking.
MI_IVE_AlphaBlending	Execute an independent weighted addition calculation task against two grayscale images.
MI_IVE_BGBlur	Execute the task of blur, replacement, mosaic or blur with mosaic for image background.

Note:

Stride
- MI_IVE_Image_t The stride of image data indicates the number of units which is calculated on the basis of "pixels" of an image in one line, the bit width of "pixel" can be 8/16-bit.
- MI_IVE_Data_t The stride of two-dimensional data indicates the number of bytes in one line of two-dimensional data.
  
  MI_IVE_Data_t can be looked upon as an image with "pixels" presented in 8-bit addressing. In this light, the stride can be interpreted as the number of units which is calculated on the basis of "pixels" of an image or two-dimensional data in one line.
  
  Fig 1-1: Data Stride
Alignment

For rapidly accessing the memory start address or data across rows, Hardware requires that the memory address/stride must be a multiple of the alignment factor.
- Start address alignment for data memory
  
  The current IVE operator requires that the Input/Output start address should be 1 / 2 / 16-byte-aligned. For details, please refer to the parameter requirement in each APIs.
- Stride alignment
  
  The stride of generalized two-dimensional images, two-dimensional single-component data and one-dimensional array data must be 16-pixel-aligned.
Input/Output data type (for detailed structure definition, please refer to Chapter IVE DATA TYPE)
- Generalized two-dimensional image data
  
  For MI_IVE_Image_t, MI_IVE_SrcImage_t, and MI_IVE_DstImage_t, please refer to MI_IVE_ImageType_e for the image type.
  - Note: Currently, the width and height of the generalized two-dimensional image data for the operator input/output must be even number.
- Two-dimensional single-component data
  
  MI_IVE_Data_t, two-dimensional data in byte unit, is mainly used for DMA, etc. Depending on the image type, MI_IVE_Image_t can be converted into one or more MI_IVE_Data_t data.
- One-dimensional data
  
  MI_IVE_MemInfo_t, MI_IVE_SrcMemInfo_t, and MI_IVE_DstMemInfo_t are one-dimensional data such as histogram statistics, GMM model data, and LK optical flow corner input, etc.

Types of generalized two-dimensional image

Image Type	Image Description	Memory Address Usage	Stride Usage
E_MI_IVE_IMAGE_TYPE_U8C1	8-bit unsigned single channel image	aphyPhyAddr[0] and apu8VirAddr[0] are used	Only u16Stride[0] is used
E_MI_IVE_IMAGE_TYPE_S8C1	8-bit signed single channel image	aphyPhyAddr[0] and apu8VirAddr[0] are used	Only u16Stride[0] is used
E_MI_IVE_IMAGE_TYPE_YUV420SP	Image in YUV420 Semiplanar format	aphyPhyAddr[0], apu8VirAddr[0] (Y component), aphyPhyAddr[1], apu8VirAddr[1] (UV component) are used the UV component data is placed in order of UV	au16Stride[0] (stride of Y component), au16Stride[1] (stride of UV component) are used
E_MI_IVE_IMAGE_TYPE_YUV422SP	Image in YUV422 Semiplanar format	aphyPhyAddr[0], apu8VirAddr[0] (Y component), aphyPhyAddr[1], apu8VirAddr[1] (UV component) are used the UV component data is placed in order of UV	au16Stride[0] (stride of Y component), au16Stride[1] (stride of UV component) are used
E_MI_IVE_IMAGE_TYPE_YUV420P	Image in YUV420 Planar format	aphyPhyAddr[0], apu8VirAddr[0] (Y component), aphyPhyAddr[1], apu8VirAddr[1] (U component), aphyPhyAddr[2], apu8VirAddr[2] (V component) are used	au16Stride[0] (stride of Y component), au16Stride[1] (stride of U component), au16Stride[2] (stride of V component) are used
E_MI_IVE_IMAGE_TYPE_YUV422P	Image in YUV422 Planar format	aphyPhyAddr[0], apu8VirAddr[0] (Y component), aphyPhyAddr[1], apu8VirAddr[1] (U component), aphyPhyAddr[2], apu8VirAddr[2] (V component) are used	au16Stride[0] (stride of Y component), au16Stride[1] (stride of U component), au16Stride[2] (stride of V component) are used
E_MI_IVE_IMAGE_TYPE_S8C2_PACKAGE	8-bit signed dual-channel image dual-channel image in package format	aphyPhyAddr[0], apu8VirAddr[0] (channel 0&1 component) are used	au16Stride[0] (stride of channel 0&1 component) is used
E_MI_IVE_IMAGE_TYPE_S8C2_PLANAR	8 bit signed dual-channel image dual-channel image in planar format	aphyPhyAddr[0], apu8VirAddr[0] (channel 0 component), aphyPhyAddr[1], apu8VirAddr[1] (channel 1 component) are used	au16Stride[0] (stride of channel 0 component), au16Stride[1] (stride of channel 1 component) are used
E_MI_IVE_IMAGE_TYPE_S16C1	16 bit signed single channel image	aphyPhyAddr[0] and apu8VirAddr[0] are used	Only u16Stride[0] is used
E_MI_IVE_IMAGE_TYPE_U16C1	16 bit unsigned single channel image	aphyPhyAddr[0] and apu8VirAddr[0] are used	Only u16Stride[0] is used
E_MI_IVE_IMAGE_TYPE_U8C3_PACKAGE	8-bit unsigned 3-channel image 3-channel image in package format	aphyPhyAddr[0], apu8VirAddr[0] (channel 0&1&2 component) are used	au16Stride[0] (stride of channel 0&1&2 component) is used
E_MI_IVE_IMAGE_TYPE_U8C3_PLANAR	8 bit unsigned 3-channel image 3-channel image in planar format	aphyPhyAddr[0], apu8VirAddr[0] (channel 0 component), aphyPhyAddr[1], apu8VirAddr[1] (channel 1 component), aphyPhyAddr[2], apu8VirAddr[2] (channel 2 component) are used	au16Stride[0] (stride of channel 0 component), au16Stride[1] (stride of channel 1 component), au16Stride[2] (stride of channel 2 component) are used
E_MI_IVE_IMAGE_TYPE_S32C1	32 bit signed single channel image	aphyPhyAddr[0] and apu8VirAddr[0] are used	Only u16Stride[0] is used
E_MI_IVE_IMAGE_TYPE_U32C1	32 bit unsigned single channel image	aphyPhyAddr[0] and apu8VirAddr[0] are used	Only u16Stride[0] is used
E_MI_IVE_IMAGE_TYPE_S64C1	64 bit signed single channel image	aphyPhyAddr[0] and apu8VirAddr[0] are used	Only u16Stride[0] is used
E_MI_IVE_IMAGE_TYPE_U64C1	64 bit unsigned single channel image	aphyPhyAddr[0] and apu8VirAddr[0] are used	Only u16Stride[0] is used
E_MI_IVE_IMAGE_TYPE_YUV420SP_NV21	Image in YUV420 Semiplanar format	aphyPhyAddr[0], apu8VirAddr[0] (Y component), aphyPhyAddr[1], apu8VirAddr[1] (UV component) are used the UV component data is placed in order of VU	au16Stride[0] (stride of Y component), au16Stride[1] (stride of UV component) are used
E_MI_IVE_IMAGE_TYPE_YUV422_YUYV	Image in YUV422 package format	Only aphyPhyAddr[0] and apu8VirAddr[0] are used (Y&U&V component) the YUV component data is placed in order of YUYV	Only u16Stride[0] is used (stride of Y&U&V component)

The matrix storage.
- Row-major: the elements in the same row are placed in order in the memory.
- Column-major: the elements in the same column are placed in order in the memory.
- The storage example of the matrix which placed in row/column-major is as follows:

2.1. MI_IVE_Create_Handle¶

Function

Create IVE handle and available handle index.
Syntax
```
MI_IVE_HANDLE MI_IVE_Create_Handle();
```
Return Value

Return Value Description

Handle index Range: [0, MI_IVE_HANDLE_MAX)

MI_IVE_HANDLE_MAX Failed.
Requirement
- Header: mi_common_datatype.h, mi_ive.h, mi_ive_datatype.h
- Library: libmi_ive.a
Note
- Please choose your process between MI_IVE_Create_Handle and MI_IVE_Create.

2.2. MI_IVE_Create¶

Function

Create an IVE handle

Syntax

MI_S32 MI_IVE_Create(MI_IVE_HANDLE hHandle);

Parameter

Parameter Name Description Input/Output

hHandle Regional handle number.
Must be an unused hHandle number.
Parameter range: [0, MI_IVE_HANDLE_MAX) Input
Return Value

Return Value Description

0 Successful.

Non-zero Failed. Please refer to ERROR CODE.
Requirement
- Header: mi_common_datatype.h, mi_ive.h, mi_ive_datatype.h
- Library: libmi_ive.a
Note
- Please choose your process between MI_IVE_Create_Handle and MI_IVE_Create.

2.3. MI_IVE_Destroy¶

Function

Release an IVE handle

Syntax

MI_S32 MI_IVE_Destroy(MI_IVE_HANDLE hHandle);

Parameter

Parameter Name Description Input/Output

hHandle Regional handle number.
Parameter range: [0, MI_IVE_HANDLE_MAX). Input
Return Value

Return Value Description

0 Successful.

Non-zero Failed. Please refer to ERROR CODE.
Requirement
- Header: mi_common_datatype.h, mi_ive.h, mi_ive_datatype.h
- Library: libmi_ive.a

2.4. MI_IVE_Filter¶

Function

Execute a 5x5 template filter task, and different filtering result including mean filtering and Gaussian filtering can be achieved by configuring specific coefficients.

Syntax

MI_S32 MI_IVE_Filter(MI_IVE_HANDLE hHandle, MI_IVE_SrcImage_t *pstSrc, MI_IVE_DstImage_t *pstDst, MI_IVE_FilterCtrl_t *pstFltCtrl, MI_BOOL bInstant);

Parameter

Parameter Name	Description	Input/Output
hHandle	Regional handle number. Parameter range: [0, MI_IVE_HANDLE_MAX).	Input
pstSrc	Source image pointer. Cannot be null.	Input
pstDst	Output image pointer. Cannot be null. Width and height same as pstSrc.	Output
pstFltCtrl	Control info pointer. Cannot be null.	Input
bInstant	Reserved.	Input

Parameter Name	Supported Image Type	Address Alignment	Resolution
pstSrc	U8C1, YUV420SP, YUV422SP	16 byte	64x64 ~ 1920x1080
pstDst	Same as pstSrc	16 byte	Same as pstSrc

U8C1, YUV420SP, and YUV422SP are the short form of MI_IVE_ImageType_e member. Other members will adopt the same naming rule throughout the rest of the document.
For Pcupid series: The hardware acceleration is supported when the format is U8C1 and the resolution is not exceeded 640x480.

Return Value

Return Value Description

0 Successful.

Non-zero Failed. Please refer to ERROR CODE.
Requirement
- Header: mi_common_datatype.h, mi_ive.h, mi_ive_datatype.h
- Library: libmi_ive.a
Note
- When the source data type is YUV420SP or YUV422SP, the output data stride must be consistent.
  
  Fig 1-2: Filter Calculation Formula.
- The calculation formula of Filter is as follows:
  
  $I_{out}(x,y)={\sum_{-2\leqslant i\leqslant2} \sum_{-2\leqslant j\leqslant2} I(x+i,y+j)*coef(x+i,y+j)}>>norm$
  
  Where, $I(x,y)$ refers to pstSrc, $I_{out}(x,y)$ refers to pstDst, $coef(x,y)$ refers to as8Mask[MI_IVE_MASK_SIZE_5X5] in pstFltCtrl, and $norm$ refers to u8Norm in pstFltCtrl.
- The classic Gaussian template is as illustrated below:
  
  $\begin{bmatrix} 0 &0 &0 &0 &0 \\ 0 &1 &2 &1 &0 \\ 0 &2 &4 &2 &0 \\ 0 &1 &2 &1 &0 \\ 0 &0 &0 &0 &0 \end{bmatrix} \ u8Norm=4 \quad \begin{bmatrix} 1 &2 &3 &2 &1 \\ 2 &5 &6 &5 &2 \\ 3 &6 &8 &6 &3 \\ 2 &5 &6 &5 &2 \\ 1 &2 &3 &2 &1 \end{bmatrix} * 3 \ \ u8Norm=8 \quad \begin{bmatrix} 1 &4 &7 &4 &1 \\ 4 &16 &26 &16 &4 \\ 7 &26 &41 &26 &7 \\ 4 &16 &26 &16 &4 \\ 1 &4 &7 &4 &1 \end{bmatrix} \ u8Norm=8$
Related API

MI_IVE_FilterAndCsc

MI_IVE_OrdStatFilter

Example

MI_S32 Sample_Filter()
{
    MI_IVE_HANDLE handle = 0;
    MI_IVE_SrcImage_t stSrc = {0};
    MI_IVE_DstImage_t stDst = {0};
    MI_IVE_FilterCtrl_t stCtrl =
    {
        // 3x3 coefficients for image sharpening
        .as8Mask =
        {
            0,0,0,0,0，
            0,-1,-1,-1,0,
            0,-1, 9,-1,0,
            0,-1,-1,-1,0,
            0,0,0,0,0,
        },
        .u8Norm = 0
        /*
        // 5x5 coefficients for mean filter
        .as8Mask =
        {
            41,41,41,41,41,
            41,41,41,41,41,
            41,41,40,41,41,
            41,41,41,41,41,
            41,41,41,41,41,
        },
        .u8Norm = 10
        */
    };

    // create IVE handle
    if (MI_SUCCESS != (s32Ret = MI_IVE_Create(handle)))
    {
        printf("Could not create IVE handle\n");
        return E_MI_ERR_FAILED;
    }

    while (loop_condition == true)
    {
        // ...set input image info to stSrc. e.g. stSrc.eType = E_MI_IVE_IMAGE_TYPE_U8C1;
        ...

        // ...set output image info to stDst. e.g. stDst.eType = E_MI_IVE_IMAGE_TYPE_U8C1;
        ...

        // Filter task.
        if (MI_SUCCESS != (s32Ret = MI_IVE_Filter(handle, &stSrc, &stDst, &stCtrl, 0)))
        {
            printf("MI_IVE_Filter() return ERROR 0x%X\n", s32Ret);
            MI_IVE_Destroy(handle);
            return E_MI_ERR_FAILED;
        }

        // get output from stDst and do something expected.
        ...
    }

    // destroy IVE handle
    MI_IVE_Destroy(handle);
    return MI_SUCCESS;
}

2.5. MI_IVE_Csc¶

Function

Execute a color space conversion task for color space conversion with respect to YUV2RGB/YUV2BGR/RGB2YUV/BGR2YUV.

Syntax

MI_S32 MI_IVE_Csc(MI_IVE_HANDLE hHandle, MI_IVE_SrcImage_t *pstSrc, MI_IVE_DstImage_t *pstDst, MI_IVE_CscCtrl_t *pstCscCtrl, MI_BOOL bInstant);

Parameter

Parameter Name	Description	Input/Output
hHandle	Regional handle number. Parameter range: [0, MI_IVE_HANDLE_MAX).	Input
pstSrc	Source image pointer. Cannot be null.	Input
pstDst	Output image pointer. Cannot be null. Width and height same as pstSrc.	Output
pstCscCtrl	Control info pointer. Cannot be null.	Input
bInstant	Reserved.	Input

Parameter Name	Supported Image Type	Address Alignment	Resolution
pstSrc	YUV420SP(NV12), YUV422SP, U8C3_PLANAR, U8C3_PACKAGE, YUV420SP_NV21	16 byte	64x64 ~ 1920x1080
pstDst	YUV420SP(NV12), YUV422SP, U8C3_PLANAR, U8C3_PACKAGE, YUV420SP_NV21	16 byte	Same as pstSrc

Return Value

Return Value Description

0 Successful.

Non-zero Failed. Please refer to ERROR CODE.
Requirement
- Header: mi_common_datatype.h, mi_ive.h, mi_ive_datatype.h
- Library: libmi_ive.a
Note
- When the output data type is U8C3_PLANAR, YUV420SP or YUV422SP, the output data stride must be consistent.
- Different work modes have different output parameter ranges. For details, please refer to MI_IVE_CscMode_e.
- Conversion formula as below :
  
  $\begin{bmatrix} Y \\ U \\ V \\ \end{bmatrix} = \begin{bmatrix} 0.299 &0.587 &0.114 \\ -0.169 &-0.331 &0.5 \\ 0.5 &-0.419 &-0.081 \\ \end{bmatrix} \begin{bmatrix} R \\ G \\ B \\ \end{bmatrix} + \begin{bmatrix} 0 \\ 128 \\ 128 \\ \end{bmatrix}$
  
  $\begin{bmatrix} R \\ G \\ B \\ \end{bmatrix} = \begin{bmatrix} 1 &-0.00093 &1.401687 \\ 1 &-0.3437 &-0.71417 \\ 1 &1.77216 &0.00099 \\ \end{bmatrix} \begin{bmatrix} Y \\ U-128 \\ V-128 \\ \end{bmatrix}$
Related API

MI_IVE_FilterAndCsc

2.6. MI_IVE_FilterAndCsc¶

Function

Execute a composite task of 5x5 template filter plus YUV2RGB color space conversion, to accomplish two functions by one single execution.

Syntax

MI_S32 MI_IVE_FilterAndCsc(MI_IVE_HANDLE hHandle, MI_IVE_SrcImage_t *pstSrc, MI_IVE_DstImage_t *pstDst, MI_IVE_FilterAndCscCtrl_t *pstFltCscCtrl, MI_BOOL bInstant);

Parameter

Parameter Name	Description	Input/Output
hHandle	Regional handle number. Parameter range: [0, MI_IVE_HANDLE_MAX).	Input
pstSrc	Source image pointer. Cannot be null.	Input
pstDst	Input image pointer. Cannot be null. Width and height same as pstSrc.	Output
pstFltCscCtrl	Control info pointer. Cannot be null.	Input
bInstant	Reserved.	Input

Parameter Name	Supported Image Type	Address Alignment	Resolution
pstSrc	YUV420SP, YUV422SP	16 byte	64x64 ~ 1920x1080
pstDst	U8C3_PLANAR, U8C3_PACKAGE	16 byte	Same as pstSrc

Return Value

Return Value Description

0 Successful.

Non-zero Failed. Please refer to ERROR CODE.
Requirement
- Header: mi_common_datatype.h, mi_ive.h, mi_ive_datatype.h
- Library: libmi_ive.a
Note
- When the output data type is U8C3_PLANAR, the output data stride must be consistent.
- Only the four work modes of YUV2RGB are supported. For details, please refer to MI_IVE_CscMode_e.
Related API

MI_IVE_Filter

2.7. MI_IVE_Sobel¶

Function

Execute a 5x5 template Sobel-like gradient calculation task

Syntax

MI_S32 MI_IVE_Sobel(MI_IVE_HANDLE hHandle, MI_IVE_SrcImage_t *pstSrc, MI_IVE_DstImage_t *pstDstH, MI_IVE_DstImage_t *pstDstV, MI_IVE_SobelCtrl_t *pstSobelCtrl, MI_BOOL bInstant);

Parameter

Parameter Name	Description	Input/Output
hHandle	Regional handle number. Parameter range: [0, MI_IVE_HANDLE_MAX).	Input
pstSrc	Source image pointer. Cannot be null.	Input
pstDstH	Gradient component image H pointer gained by template filtering. According to pstSobelCtrl→eOutCtrl, this parameter cannot be null if output is required. Width and height same as pstSrc.	Output
pstDstV	Gradient component image V pointer gained by transposed template filtering. According to pstSobelCtrl→eOutCtrl, this parameter cannot be null if output is required. Width and height same as pstSrc.	Output
pstSobelCtrl	Control info pointer. Cannot be null.	Input
bInstant	Reserved.	Input

Parameter Name	Supported Image Type	Address Alignment	Resolution
pstSrc	U8C1	16 byte	64x64 ~ 1920x1080
pstDstH	S16C1	16 byte	Same as pstSrc
pstDstV	S16C1	16 byte	Same as pstSrc

Return Value

Return Value Description

0 Successful.

Non-zero Failed. Please refer to ERROR CODE.
Requirement
- Header: mi_common_datatype.h, mi_ive.h, mi_ive_datatype.h
- Library: libmi_ive.a
Note
- When the output mode is E_MI_IVE_SOBEL_OUT_CTRL_BOTH, the stride of pstDstH and the stride of pstDstV must be consistent.
- Three output modes are available for configuration. Please refer to MI_IVE_SobelOutCtrl_e.
  
  Fig 1-3: Sobel Calculation Formula.
- The calculation formula of Sobel is as follows:
  
  $H_{out}(x,y)=\sum_{-2\leqslant i\leqslant2}\sum_{-2\leqslant j\leqslant2}I(x+i,y+j)*coef(x+i,y+j)$
  
  $V_{out}(x,y)=\sum_{-2\leqslant i\leqslant2}\sum_{-2\leqslant j\leqslant2}I(x+i,y+j)*coef(x+j,y+i)$
  
  Where, $I(x, y)$ refers to pstSrc, $H_{out}(x, y)$ refers to pstDstH, $V_{out}(x, y)$ refers to pstDstV, and $coef(x,y)$ refers to as8Mask[MI_IVE_MASK_SIZE_5X5] in pstSobelCtrl.
- Sobel template
  
  $\begin{bmatrix} 0 &0 &0 &0 &0 \\ 0 &-1 &0 &1 &0 \\ 0 &-2 &0 &2 &0 \\ 0 &-1 &0 &1 &0 \\ 0 &0 &0 &0 &0 \end{bmatrix} \begin{bmatrix} 0 &0 &0 &0 &0 \\ 0 &-1 &-2 &-1 &0 \\ 0 &0 &0 &0 &0 \\ 0 &1 &2 &1 &0 \\ 0 &0 &0 &0 &0 \end{bmatrix}$
  
  $\begin{bmatrix} -1 &-2 &0 &2 &1 \\ -4 &-8 &0 &8 &4 \\ -6 &-12 &0 &12 &6 \\ -4 &-8 &0 &8 &4 \\ -1 &-2 &0 &2 &1 \end{bmatrix} \begin{bmatrix} -1 &-4 &-6 &-4 &-1 \\ -2 &-8 &-12 &-8 &-2 \\ 0 &0 &0 &0 &0 \\ 2 &8 &12 &8 &2 \\ 1 &4 &6 &4 &1 \end{bmatrix}$
- Scharr template
  
  $\begin{bmatrix} 0 &0 &0 &0 &0 \\ 0 &-3 &0 &3 &0 \\ 0 &-10 &0 &10 &0 \\ 0 &-3 &0 &3 &0 \\ 0 &0 &0 &0 &0 \end{bmatrix} \begin{bmatrix} 0 &0 &0 &0 &0 \\ 0 &-3 &-10 &-3 &0 \\ 0 &0 &0 &0 &0 \\ 0 &3 &10 &3 &0 \\ 0 &0 &0 &0 &0 \end{bmatrix}$
- Laplace template
  
  $\begin{bmatrix} 0 &0 &0 &0 &0 \\ 0 &0 &1 &0 &0 \\ 0 &1 &-4 &1 &0 \\ 0 &0 &1 &0 &0 \\ 0 &0 &0 &0 &0 \end{bmatrix} \begin{bmatrix} 0 &0 &0 &0 &0 \\ 0 &0 &-1 &0 &0 \\ 0 &-1 &4 &-1 &0 \\ 0 &0 &-1 &0 &0 \\ 0 &0 &0 &0 &0 \end{bmatrix}$
  
  $\begin{bmatrix} 0 &0 &0 &0 &0 \\ 0 &1 &1 &1 &0 \\ 0 &1 &-8 &1 &0 \\ 0 &1 &1 &1 &0 \\ 0 &0 &0 &0 &0 \end{bmatrix} \begin{bmatrix} 0 &0 &0 &0 &0 \\ 0 &-1 &-1 &-1 &0 \\ 0 &-1 &8 &-1 &0 \\ 0 &-1 &-1 &-1 &0 \\ 0 &0 &0 &0 &0 \end{bmatrix}$
Related API

MI_IVE_MagAndAng

MI_IVE_NormGrad

2.8. MI_IVE_MagAndAng¶

Function

Execute a 5x5 template gradient magnitude and angle calculation task.

Syntax

MI_S32 MI_IVE_MagAndAng(MI_IVE_HANDLE hHandle, MI_IVE_SrcImage_t *pstSrc, MI_IVE_DstImage_t *pstDstMag, MI_IVE_DstImage_t *pstDstAng, MI_IVE_MagAndAngCtrl_t *pstMagAndAngCtrl, MI_BOOL bInstant);

Parameter

Parameter Name	Description	Input/Output
hHandle	Regional handle number. Parameter range: [0, MI_IVE_HANDLE_MAX).	Input
pstSrc	Source image pointer. Cannot be null.	Input
pstDstMag	Output magnitude image pointer. Cannot be null. Width and height same as pstSrc.	Output
pstDstAng	Output angle image pointer. According to pstMagAndAngCtrl→eOutCtrl, this parameter cannot be null if output is required. Width and height same as pstSrc.	Output
pstMagAndAngCtrl	Control info pointer. Cannot be null.	Input
bInstant	Reserved.	Input

Parameter Name	Supported Image Type	Address Alignment	Resolution
pstSrc	U8C1	16 byte	64x64 ~ 1920x1080
pstDstMag	U16C1	16 byte	Same as pstSrc
pstDstAng	U8C1	16 byte	Same as pstSrc

Return Value

Return Value Description

0 Successful.

Non-zero Failed. Please refer to ERROR CODE.
Requirement
- Header: mi_common_datatype.h, mi_ive.h, mi_ive_datatype.h
- Library: libmi_ive.a
Note
- When the output mode is E_MI_IVE_MAG_AND_ANG_OUT_CTRL_MAG_AND_ANG, the stride of pstDstMag and the stride of pstDstAng must be consistent.
- Two output modes can be configured. For details, please refer to MI_IVE_MagAndAngOutCtrl_e.
  
  Fig 1-4: MagAndAng Calculation Formula.
- User can utilize pstMagAndAngCtrl→u16Thr to do thresholding operation against magnitude map (to achieve EOH).
- The calculation formula of MagAndAnd is as follows:
  
  $H_{out}(x,y)=\sum_{-2\leqslant i\leqslant2}\sum_{-2\leqslant j\leqslant2}I(x+i,y+j)*coef(x+i,y+j)$
  
  $V_{out}(x,y)=\sum_{-2\leqslant i\leqslant2}\sum_{-2\leqslant j\leqslant2}I(x+i,y+j)*coef(x+j,y+i)$
  
  $Mag(x,y)=abs(H_{out}(x,y))+abs(V_{out}(x,y))$
  
  $Mag(x,y)=\left\{ \begin{aligned} 0 , Mag(x,y) < u16Thr \\ Mag(x,y) ,Mag(x,y)\geq u16Thr \end{aligned} \right.$
  
  Where,
  - $I(x,y)$ refers to pstSrc, $Mag(x,y)$ refers to pstDstMag, $\theta(x,y)$ refers to pstDstAng, and $coef(x,y)$ refers to as8Mask[MI_IVE_MASK_SIZE_5X5] in pstMagAndAngCtrl.
  - According to $H_{out}(x,y)$ , $V_{out}(x,y)$ and $arctan(V_{out}/H_{out})$ , $\theta(x,y)$ takes the direction value corresponding to 0 ~ 7 in the figure below.
Related API

MI_IVE_CannyHysEdge

MI_IVE_CannyEdge

MI_IVE_Sobel

2.9. MI_IVE_Dilate¶

Function

Execute a binary image 5x5 template dilating task.

Syntax

MI_S32 MI_IVE_Dilate(MI_IVE_HANDLE hHandle, MI_IVE_SrcImage_t *pstSrc, MI_IVE_DstImage_t *pstDst, MI_IVE_DilateCtrl_t *pstDilateCtrl, MI_BOOL bInstant);

Parameter

Parameter Name	Description	Input/Output
hHandle	Regional handle number. Parameter range: [0, MI_IVE_HANDLE_MAX).	Input
pstSrc	Source image pointer. Cannot be null.	Input
pstDst	Output image pointer. Cannot be null. Width and height same as pstSrc.	Output
pstDilateCtrl	Control info pointer.	Input
bInstant	Reserved.	Input

Parameter Name	Supported Image Type	Address Alignment	Resolution
pstSrc	U8C1 binary image	16 byte	64x64 ~ 1920x1080
pstDst	U8C1 binary image	16 byte	Same as pstSrc

For Pcupid series: The hardware acceleration is supported when the resolution is not exceeded 640x480.

Return Value

Return Value Description

0 Successful.

Non-zero Failed. Please refer to ERROR CODE.
Requirement
- Header: mi_common_datatype.h, mi_ive.h, mi_ive_datatype.h
- Library: libmi_ive.a
Note
- The template coefficient can only be 0 or 255.
  
  Fig 1-5: Dilate Calculation Formula
- The calculation formula of Dilate is as follows:
  
  $I_{out}(x,y)=O_{-2\leqslant i\leqslant2}(O_{-2\leqslant j\leqslant2}(f(i,j))) \\$
  
  $O_{-2\leqslant k\leqslant 2}(g(k))=g(-2)\ |\ g(-1)\ |\ g(0)\ |\ g(1)\ |\ g(2) \\$
  
  $f(x,y)=I(x,y)\ \& \ coef(x,y)$
  
  In the formula | is a bitwise OR operation, & is a bitwise AND operation, and % is a remainder operation. $I(x,y)$ refers to pstSrc, $I_{out}(x,y)$ refers to pstDst, and $coef(x,y)$ refers to au8Mask[MI_IVE_MASK_SIZE_5X5] in pstDilateCtrl.
- Template example
  
  $\begin{bmatrix} 0 &0 &0 &0 &0 \\ 0 &0 &255 &0 &0 \\ 0 &255 &255 &255 &0 \\ 0 &0 &255 &0 &0 \\ 0 &0 &0 &0 &0 \end{bmatrix} \begin{bmatrix} 0 &0 &0 &0 &0 \\ 0 &255 &255 &255 &0 \\ 0 &255 &255 &255 &0 \\ 0 &255 &255 &255 &0 \\ 0 &0 &0 &0 &0 \end{bmatrix}$
  
  $\begin{bmatrix} 0 &255 &255 &255 &0 \\ 255 &255 &255 &255 &255 \\ 255 &255 &255 &255 &255 \\ 255 &255 &255 &255 &255 \\ 0 &255 &255 &255 &0 \end{bmatrix} \begin{bmatrix} 255 &255 &255 &255 &255 \\ 255 &255 &255 &255 &255 \\ 255 &255 &255 &255 &255 \\ 255 &255 &255 &255 &255 \\ 255 &255 &255 &255 &255 \end{bmatrix}$
Related API

MI_IVE_Erode

MI_IVE_OrdStatFilter

2.10. MI_IVE_Erode¶

Function

Execute a binary image 5x5 template erosion task.

Syntax

MI_S32 MI_IVE_Erode(MI_IVE_HANDLE hHandle, MI_IVE_SrcImage_t *pstSrc, MI_IVE_DstImage_t *pstDst, MI_IVE_ErodeCtrl_t *pstErodeCtrl, MI_BOOL bInstant);

Parameter

Parameter Name	Description	Input/Output
hHandle	Regional handle number. Parameter range: [0, MI_IVE_HANDLE_MAX).	Input
pstSrc	Source image pointer. Cannot be null.	Input
pstDst	Output image pointer. Cannot be null. Width and height same as pstSrc.	Output
pstErodeCtrl	Control info pointer. Cannot be null.	Input
bInstant	Reserved.	Input

Parameter Name	Supported Image Type	Address Alignment	Resolution
pstSrc	U8C1 binary image	16 byte	64x64 ~ 1920x1080
pstDst	U8C1 binary image	16 byte	Same as pstSrc

For Pcupid series: The hardware acceleration is supported when the resolution is not exceeded 640x480.

Return Value

Return Value Description

0 Successful.

Non-zero Failed. Please refer to ERROR CODE.
Requirement
- Header: mi_common_datatype.h, mi_ive.h, mi_ive_datatype.h
- Library: libmi_ive.a
Note
- The template coefficient can only be 0 or 255.
  
  Fig 1-6: Erode Calculation Formula
- The calculation formula of Erode is as follows:
  
  $I_{out}(x,y)=O_{-2\leqslant i\leqslant 2}(O_{-2\leqslant j\leqslant2}(f(x+i,y+j)))$
  
  $O_{-2\leqslant k\leqslant 2}(g(k))=g(-2)\ |\ g(-1)\ |\ g(0)\ |\ g(1)\ |\ g(2)$
  
  $f(x,y) = I(x,y)\ | \ (255 - coef(x,y))$
  
  In the formula | is a bitwise OR operation, & is a bitwise AND operation, and % is a remainder operation. $I(x, y)$ refers to pstSrc, $I_{out}(x, y)$ refers to pstDst, and $coef(x,y)$ refers to au8Mask[MI_IVE_MASK_SIZE_5X5] in pstErodeCtrl.
- Template example
  
  $\begin{bmatrix} 0 &0 &0 &0 &0 \\ 0 &0 &255 &0 &0 \\ 0 &255 &255 &255 &0 \\ 0 &0 &255 &0 &0 \\ 0 &0 &0 &0 &0 \end{bmatrix} \begin{bmatrix} 0 &0 &0 &0 &0 \\ 0 &255 &255 &255 &0 \\ 0 &255 &255 &255 &0 \\ 0 &255 &255 &255 &0 \\ 0 &0 &0 &0 &0 \end{bmatrix}$
  
  $\begin{bmatrix} 0 &255 &255 &255 &0 \\ 255 &255 &255 &255 &255 \\ 255 &255 &255 &255 &255 \\ 255 &255 &255 &255 &255 \\ 0 &255 &255 &255 &0 \end{bmatrix} \begin{bmatrix} 255 &255 &255 &255 &255 \\ 255 &255 &255 &255 &255 \\ 255 &255 &255 &255 &255 \\ 255 &255 &255 &255 &255 \\ 255 &255 &255 &255 &255 \end{bmatrix}$
Related API

MI_IVE_Dilate

MI_IVE_OrdStatFilter

2.11. MI_IVE_Thresh¶

Function

Execute a grayscale image thresholding task.

Syntax

MI_S32 MI_IVE_Thresh(MI_IVE_HANDLE hHandle, MI_IVE_SrcImage_t *pstSrc, MI_IVE_DstImage_t *pstDst, MI_IVE_ThreshCtrl_t *pstThrCtrl, MI_BOOL bInstant);

Parameter

Parameter Name	Description	Input/Output
hHandle	Regional handle number. Parameter range: [0, MI_IVE_HANDLE_MAX).	Input
pstSrc	Source image pointer. Cannot be null.	Input
pstDst	Output image pointer. Cannot be null. Width and height same as pstSrc.	Output
pstThrCtrl	Control info pointer.	Input
bInstant	Reserved.	Input

Parameter Name	Supported Image Type	Address Alignment	Resolution
pstSrc	U8C1	1 byte	64x64 ~ 1920x1080
pstDst	U8C1	1 byte	Same as pstSrc

Return Value

Return Value Description

0 Successful.

Non-zero Failed. Please refer to ERROR CODE.
Requirement
- Header: mi_common_datatype.h, mi_ive.h, mi_ive_datatype.h
- Library: libmi_ive.a
Note
- Eight operation modes are available for configuration. For details, please refer to MI_IVE_ThreshMode_e.
  
  Fig 1-7: Threshold Modes of Thresh.
- The calculation formulas of different mode are as follows:
  - E_MI_IVE_THRESH_MODE_BINARY:
    
    $I_{out}(x,y)=\left\{ \begin{aligned} minVal,\ I(x,y)\leqslant lowThr \\ maxVal,\ I(x,y)> lowThr \end{aligned} \right.$
    
    midVal, highThr do not require to be assigned.
  - E_MI_IVE_THRESH_MODE_TRUNC:
    
    $I_{out}(x,y)=\left\{ \begin{aligned} I(x,y),\ I(x,y)\leqslant lowThr \\ maxVal,\ I(x,y)>lowThr \end{aligned} \right.$
    
    minVal, midVal, highThr do not require to be assigned.
  - E_MI_IVE_THRESH_MODE_TO_MINVAL:
    
    $I_{out}(x,y)=\left\{ \begin{aligned} minVal,\ I(x,y)\leqslant lowThr \\ I(x,y),\ I(x,y)> lowThr \end{aligned} \right.$
    
    midVal, maxVal, highThr do not require to be assigned.
  - E_MI_IVE_THRESH_MODE_MIN_MID_MAX:
    
    $I_{out}(x,y)=\left\{ \begin{aligned} minVal,\ I(x,y)\leqslant lowThr \\ midVal,\ lowThr\leqslant I(x,y)\leqslant highThr \\ maxVal,\ I(x,y)>highThr \end{aligned} \right.$
  - E_MI_IVE_THRESH_MODE_ORI_MID_MAX:
    
    $I_{out}(x,y)=\left\{ \begin{aligned} I(x,y),\ I(x,y)\leqslant lowThr \\ midVal,\ lowThr\leqslant I(x,y)\leqslant highThr \\ maxVal,\ I(x,y)>highThr \end{aligned} \right.$
    
    minVal does not require to be assigned.
  - E_MI_IVE_THRESH_MODE_MIN_MID_ORI:
    
    $I_{out}(x,y)=\left\{ \begin{aligned} minVal,\ I(x,y) \leqslant lowThr \\ midVal,\ lowThr\leqslant I(x,y)\leqslant highThr \\ I(x,y),\ I(x,y)>highThr \end{aligned} \right.$
    
    maxVal does not require to be assigned.
  - E_MI_IVE_THRESH_MODE_MIN_ORI_MAX:
    
    $I_{out}(x,y)=\left\{ \begin{aligned} minVal,\ I(x,y) \leqslant lowThr \\ I(x,y),\ lowThr\leqslant I(x,y)\leqslant highThr \\ maxVal,\ I(x,y)>highThr \end{aligned} \right.$
    
    midVal does not require to be assigned.
  - E_MI_IVE_THRESH_MODE_ORI_MID_ORI:
    
    $I_{out}(x,y)=\left\{ \begin{aligned} I(x,y),\ I(x,y) \leqslant lowThr \\ midVal,\ lowThr\leqslant I(x,y)\leqslant highThr \\ I(x,y),\ I(x,y)>highThr \end{aligned} \right.$
    
    minVal, maxVal do not require to be assigned
    
    Where,
    - $I(x,y)$ refers to pstSrc, $I_{out}(x,y)$ refers to pstDst.
    - $mode$ , $lowThr$ , $highThr$ , $minVal$ , $midVal$ and $maxVal$ refer respectively to eMode, u8LowThr, u8HighThr, u8MinVal, u8MidVal and u8MaxVal in pstThrCtrl.
    - u8MinVal, u8MidVal and u8MaxVal in pstThrCtrl do not need to satisfy the size relationship signified by the variable name.
Related API

MI_IVE_ThreshS16

MI_IVE_ThreshU16

2.12. MI_IVE_And¶

Function

Execute an AND task against two binary images.

Syntax

MI_S32 MI_IVE_And(MI_IVE_HANDLE hHandle, MI_IVE_SrcImage_t *pstSrc1, MI_IVE_SrcImage_t *pstSrc2, MI_IVE_DstImage_t *pstDst, MI_BOOL bInstant);

Parameter

Parameter Name	Description	Input/Output
hHandle	Regional handle number. Parameter range: [0, MI_IVE_HANDLE_MAX).	Input
pstSrc1	Source image 1 pointer. Cannot be null.	Input
pstSrc2	Source image 2 pointer. Cannot be null. Width and height same as pstSrc1.	Input
pstDst	Output image pointer. Cannot be null. Width and height same as pstSrc1.	Output
bInstant	Reserved.	Input

Parameter Name	Supported Image Type	Address Alignment	Resolution
pstSrc1	U8C1 binary image	1 byte	64x64 ~ 1920x1080
pstSrc2	U8C1 binary image	1 byte	Same as pstSrc1
pstDst	U8C1 binary image	1 byte	Same as pstSrc1

Return Value

Return Value Description

0 Successful.

Non-zero Failed. Please refer to ERROR CODE.
Requirement
- Header: mi_common_datatype.h, mi_ive.h, mi_ive_datatype.h
- Library: libmi_ive.a
Note
- The calculation formula of And is as follows:
  
  $I_{out}(x,y)=I_{src1}(x,y)\ \&\ I_{src2}(x,y)$
  
  Where, $I_{src1}(x,y)$ refers to pstSrc1, $I_{src2}(x,y)$ refers to pstSrc2, and $I_{out}(x,y)$ refers to pstDst
Related API

MI_IVE_Or

MI_IVE_Xor

2.13. MI_IVE_Sub¶

Function

Execute a SUBTRACT task against two grayscale images.

Syntax

MI_S32 MI_IVE_Sub(MI_IVE_HANDLE hHandle, MI_IVE_SrcImage_t *pstSrc1, MI_IVE_SrcImage_t *pstSrc2, MI_IVE_DstImage_t *pstDst, MI_IVE_SubCtrl_t *pstSubCtrl, MI_BOOL bInstant);

Parameter

Parameter Name	Description	Input/Output
hHandle	Regional handle number. Parameter range: [0, MI_IVE_HANDLE_MAX).	Input
pstSrc1	Source image 1 pointer. Cannot be null.	Input
pstSrc2	Source image 2 pointer. Cannot be null. Width and height same as pstSrc1.	Input
pstDst	Output image pointer. Cannot be null. Width and height same as pstSrc1.	Output
pstSubCtrl	Control info pointer. Cannot be null.	Input
bInstant	Reserved.	Input

Parameter Name	Supported Image Type	Address Alignment	Resolution
pstSrc1	U8C1	1 byte	64x64 ~ 1920x1080
pstSrc2	U8C1	1 byte	Same as pstSrc1
pstDst	U8C1, S8C1	1 byte	Same as pstSrc1

Return Value

Return Value Description

0 Successful.

Non-zero Failed. Please refer to ERROR CODE.
Requirement
- Header: mi_common_datatype.h, mi_ive.h, mi_ive_datatype.h
- Library: libmi_ive.a
Note
- Two output formats are available for configuration. For details, please refer to MI_IVE_SubMode_e.
- The calculation formulas of differernt mode are as follows:
  - E_MI_IVE_SUB_MODE_ABS
    
    Calculation formula: $I_{out}(x,y)=abs(I_{src1}(x,y) - I_{src2}(x,y))$
    
    Output format: U8C1
  - E_MI_IVE_SUB_MODE_SHIFT
    
    Calculation formula: $I_{out}(x,y)=(I_{src1}(x,y) - I_{src2}(x,y))>>1$
    
    Output format: S8C1
  Where, $I_{src1}(x,y)$ refers to pstSrc1, $I_{src2}(x,y)$ refers to pstSrc2, and $I_{out}(x,y)$ refers to pstDst.
Related API

MI_IVE_Add

2.14. MI_IVE_Or¶

Function

Execute an OR task against two binary images.

Syntax

MI_S32 MI_IVE_Or(MI_IVE_HANDLE hHandle, MI_IVE_SrcImage_t *pstSrc1, MI_IVE_SrcImage_t *pstSrc2, MI_IVE_DstImage_t *pstDst, MI_BOOL bInstant);

Parameter

Parameter Name	Description	Input/Output
hHandle	Regional handle number. Parameter range: [0, MI_IVE_HANDLE_MAX).	Input
pstSrc1	Source image 1 pointer. Cannot be null.	Input
pstSrc2	Source image 2 pointer. Cannot be null. Width and height same as pstSrc1.	Input
pstDst	Output image pointer. Cannot be null. Width and height same as pstSrc1.	Output
bInstant	Reserved.	Input

Parameter Name	Supported Image Type	Address Alignment	Resolution
pstSrc1	U8C1	1 byte	64x64 ~ 1920x1080
pstSrc2	U8C1	1 byte	Same as pstSrc1
pstDst	U8C1	1 byte	Same as pstSrc1

Return Value

Return Value Description

0 Successful.

Non-zero Failed. Please refer to ERROR CODE.
Requirement
- Header: mi_common_datatype.h, mi_ive.h, mi_ive_datatype.h
- Library: libmi_ive.a
Note
- The calculation formula of Or is as follows:
  
  $I_{out}(x,y)=I_{src1}(x,y)\ | \ I_{src2}(x,y)$
  
  Where, $I_{src1}(x,y)$ refers to pstSrc1, $I_{src2}(x,y)$ refers to pstSrc2, and $I_{out}(x,y)$ refers to pstDst.
Related API

MI_IVE_And

MI_IVE_Xor

2.15. MI_IVE_Integ¶

Function

Execute an integral graph statistics task against grayscale images.

Syntax

MI_S32 MI_IVE_Integ(MI_IVE_HANDLE hHandle, MI_IVE_SrcImage_t *pstSrc, MI_IVE_DstImage_t *pstDst, MI_IVE_IntegCtrl_t *pstIntegCtrl, MI_BOOL bInstant);

Parameter

Parameter Name	Description	Input/Output
hHandle	Regional handle number. Parameter range: [0, MI_IVE_HANDLE_MAX).	Input
pstSrc	Source image pointer. Cannot be null.	Input
pstDst	Output image pointer. Cannot be null. Width and height same as pstSrc.	Output
pstIntegCtrl	Control info pointer. Cannot be null.	Input
bInstant	Reserved.	Input

Parameter Name	Supported Image Type	Address Alignment	Resolution
pstSrc	U8C1	16 byte	32x16 ~ 1920x1080
pstDst	U32C1, U64C1	16 byte	Same as pstSrc

Return Value

Return Value Description

0 Successful.

Non-zero Failed. Please refer to ERROR CODE.
Requirement
- Header: mi_common_datatype.h, mi_ive.h, mi_ive_datatype.h
- Library: libmi_ive.a
Note
- The calculation formulas of different mode are as follows:
  - E_MI_IVE_INTEG_OUT_CTRL_COMBINE(the combined output mode, the output image type must be E_MI_IVE_IMAGE_TYPE_U64C1)
    
    $I_{sum}(x,y)=\sum^{i\leqslant x}_{i\geqslant0}\sum^{j\leqslant y}_{j\geqslant0}I(i,j) ,\quad I_{sq}(x,y)=\sum^{i\leqslant x}_{i\geqslant0}\sum^{j\leqslant y}_{j\geqslant0}(I(i,j)*I(i,j))$
    
    $I_{out}(x,y)=(i_{sq}(x,y)<<28)\ | \ (I_{sum}(x,y)\&0xFFFFFFF)$
  - E_MI_IVE_INTEG_OUT_CTRL_SUM(the integral graph output mode, the output image type must be E_MI_IVE_IMAGE_TYPE_U32C1)
    
    $I_{out}(x,y) = I_{sum}(x,y)=\sum^{i\leqslant x}_{i\geqslant0}\sum^{j\leqslant y}_{j\geqslant0}I(i,j)$
  - E_MI_IVE_INTEG_OUT_CTRL_SQSUM(the square and integral graph output mode, the output image type must be E_MI_IVE_IMAGE_TYPE_U64C1)
    
    $I_{out}(x,y) = I_{sq}(x,y)=\sum^{i\leqslant x}_{i\geqslant0}\sum^{j\leqslant y}_{j\geqslant0}(I(i,j)*I(i,j))$
  Where, $I(x,y)$ refers to pstSrc, and $I_{out}(x,y)$ refers to pstDst.

2.16. MI_IVE_Hist¶

Function

Execute a histogram statistics task.

Syntax

MI_S32 MI_IVE_Hist(MI_IVE_HANDLE hHandle, MI_IVE_SrcImage_t *pstSrc, MI_IVE_DstMemInfo_t *pstDst, MI_BOOL bInstant);

Parameter

Parameter Name	Description	Input/Output
hHandle	Regional handle number. Parameter range: [0, MI_IVE_HANDLE_MAX).	Input
pstSrc	Source image pointer. Cannot be null.	Input
pstDst	Output data pointer. Cannot be null. The memory should at least has 1024 bytes.	Output
bInstant	Reserved.	Input

Parameter Name	Supported Image Type	Address Alignment	Resolution
pstSrc	U8C1	16 byte	64x64 ~ 1920x1080
pstDst	-	16 byte	-

Return Value

Return Value Description

0 Successful.

Non-zero Failed. Please refer to ERROR CODE.
Requirement
- Header: mi_common_datatype.h, mi_ive.h, mi_ive_datatype.h
- Library: libmi_ive.a
Note
- The calculation formula of Hist is as follows:
  
  $I_{out}(x)=\sum_i\sum_j((I(i,j)==x)?1:0) ,x=0...255$
  
  Where, $I(i,j)$ refers to pstSrc, and $I_{out}(x)$ refers to pstDst.

2.17. MI_IVE_ThreshS16¶

Function

Execute an S16 data to 8-bit data thresholding task.

Syntax

MI_S32 MI_IVE_ThreshS16(MI_IVE_HANDLE hHandle, MI_IVE_SrcImage_t *pstSrc, MI_IVE_DstImage_t *pstDst, MI_IVE_ThreshS16Ctrl_t *pstThrS16Ctrl, MI_BOOL bInstant);

Parameter

Parameter Name	Description	Input/Output
hHandle	Regional handle number. Parameter range: [0, RGN_HANDLE_MAX).	Input
pstSrc	Source image pointer. Cannot be null.	Input
pstDst	Output image pointer. Cannot be null. Width and height same as pstSrc.	Output
pstThrS16Ctrl	Control parameter pointer. Cannot be null.	Input
bInstant	Reserved.	Input

Parameter Name	Supported Image Type	Address Alignment	Resolution
pstSrc	S16C1	2 byte	64x64 ~ 1920x1080
pstDst	U8C1, S8C1	1 byte	Same as pstSrc

Return Value

Return Value Description

0 Successful.

Non-zero Failed. Please refer to ERROR CODE.
Requirement
- Header: mi_common_datatype.h, mi_ive.h, mi_ive_datatype.h
- Library: libmi_ive.a
Note
- Four operation modes are available for configuration. For details, please refer to MI_IVE_ThreshS16Mode_e.
  
  Fig 1-8: Threshold Modes of ThreshS16
- The calculation formulas of different mode are as follows:
  - E_MI_IVE_THRESH_S16_MODE_S16_TO_S8_MIN_MID_MAX:
    
    $I_{out}(x,y)=\left \{ \begin{aligned} minVal,(I(x,y)\leqslant lowThr) \\ midVal,(lowThr< I(x,y)\leqslant highThr) \\ maxVal,(I(x,y)> highThr) \end{aligned} \right.$
    
    Requirement: $-32768 ≤ lowThr ≤ highThr ≤32767$ , $-128 ≤ minVal, midVal, maxVal ≤ 127$ .
  - E_MI_IVE_THRESH_S16_MODE_S16_TO_S8_MIN_ORI_MAX:
    
    $I_{out}(x,y)=\left \{ \begin{aligned} minVal,(I(x,y)\leqslant lowThr) \\ I(x,y),(lowThr< I(x,y)\leqslant highThr) \\ maxVal,(I(x,y)> highThr) \end{aligned} \right.$
    
    Requirement: $-128 ≤ lowThr ≤ highThr ≤127$ , $-128 ≤ minVal, maxVal ≤127$ .
  - E_MI_IVE_THRESH_S16_MODE_S16_TO_U8_MIN_MID_MAX:
    
    $I_{out}(x,y)=\left \{ \begin{aligned} minVal,(I(x,y)\leqslant lowThr) \\ midVal,(lowThr< I(x,y)\leqslant highThr) \\ maxVal,(I(x,y)> highThr) \end{aligned} \right.$
    
    Requirement: $-32768 ≤ lowThr ≤ highThr ≤32767$ , $0 ≤ minVal, midVal, maxVal ≤255$ .
  - E_MI_IVE_THRESH_S16_MODE_S16_TO_U8_MIN_ORI_MAX:
    
    $I_{out}(x,y)=\left \{ \begin{aligned} minVal,(I(x,y)\leqslant lowThr) \\ I(x,y),(lowThr< I(x,y)\leqslant highThr) \\ maxVal,(I(x,y)> highThr) \end{aligned} \right.$
    
    Requirement: $-1 ≤ lowThr ≤ highThr ≤255$ , $ 0 ≤ minVal, maxVal ≤255$ .
    
    Where,
    - $I(x,y)$ refers to pstSrc, $I_{out}(x,y)$ refers to pstDst.
    - $mode$ , $lowThr$ , $highThr$ , $minVal$ , $midVal$ and $maxVal$ refer respectively to eMode, s16LowThr, s16HighThr, un8MinVal, un8MidVal and un8MaxVal in pstThrS16Ctrl.
    - un8MinVal, un8MidVal and un8MaxVal in pstThrS16Ctrl do not need to satisfy the size relationship signified by the variable name.
Related API

MI_IVE_ThreshU16

MI_IVE_16BitTo8Bit

2.18. MI_IVE_ThreshU16¶

Function

Execute a U16 data to U8 data thresholding task.

Syntax

MI_S32 MI_IVE_ThreshU16(MI_IVE_HANDLE hHandle, MI_IVE_SrcImage_t *pstSrc, MI_IVE_DstImage_t *pstDst, MI_IVE_ThreshU16Ctrl_t *pstThrU16Ctrl, MI_BOOL bInstant);

Parameter

Parameter Name	Description	Input/Output
hHandle	Regional handle number. Parameter range: [0, MI_IVE_HANDLE_MAX).	Input
pstSrc	Source image pointer. Cannot be null.	Input
pstDst	Output image pointer. Cannot be null. Width and height same as pstSrc.	Output
pstThrU16Ctrl	Control parameter pointer. Cannot be null.	Input
bInstant	Reserved.	Input

Parameter Name	Supported Image Type	Address Alignment	Resolution
pstSrc	U16C1	2 byte	64x64 ~ 1920x1080
pstDst	U8C1	1 byte	Same as pstSrc

Return Value

Return Value Description

0 Successful.

Non-zero Failed. Please refer to ERROR CODE.
Requirement
- Header: mi_common_datatype.h, mi_ive.h, mi_ive_datatype.h
- Library: libmi_ive.a
Note
- Two operation modes are available for configuration. For details, please refer to MI_IVE_ThreshU16Mode_e.
  
  Fig 1-9: Threshold Modes of ThreshU16
- The calculation formulas of different mode are as follows:
  - E_MI_IVE_THRESH_U16_MODE_U16_TO_U8_MIN_MID_MAX:
    
    $I_{out}(x,y)=\left \{ \begin{aligned} minVal,(I(x,y)\leqslant lowThr) \\ midVal,(lowThr< I(x,y)\leqslant highThr) \\ maxVal,(I(x,y)> highThr) \end{aligned} \right.$
    
    Requirement: $0 ≤ lowThr ≤ highThr ≤65535$ ;
  - E_MI_IVE_THRESH_U16_MODE_U16_TO_U8_MIN_ORI_MAX:
    
    $I_{out}(x,y)=\left \{ \begin{aligned} minVal,(I(x,y)\leqslant lowThr) \\ I(x,y),(lowThr< I(x,y)\leqslant highThr) \\ maxVal,(I(x,y)> highThr) \end{aligned} \right.$
    
    Requirement: $0 ≤ lowThr ≤ highThr ≤255$ ;
  Where,
  - $I(x,y)$ refers to pstSrc, $I_{out}(x,y)$ refers to pstDst.
  - $mode$ , $lowThr$ , $highThr$ , $minVal$ , $midVal$ and $maxVal$ refer respectively to eMode, u16LowThr, u16HighThr, u8MinVal, u8MidVal and u8MaxVal in pstThrU16Ctrl.
  - u8MinVal, u8MidVal and u8MaxVal in pstThrU16Ctrl do not need to satisfy the size relationship signified by the variable name.
Related API

MI_IVE_ThreshS16

MI_IVE_16BitTo8Bit

2.19. MI_IVE_16BitTo8Bit¶

Function

Execute a 16-bit image data to 8-bit image data linear transformation task.

Syntax

MI_S32 MI_IVE_16BitTo8Bit(MI_IVE_HANDLE hHandle, MI_IVE_SrcImage_t *pstSrc, MI_IVE_DstImage_t *pstDst, MI_IVE_16bitTo8BitCtrl_t *pst16BitTo8BitCtrl, MI_BOOL bInstant);

Parameter

Parameter Name	Description	Input/Output
hHandle	Regional handle number. Parameter range: [0, MI_IVE_HANDLE_MAX).	Input
pstSrc	Source image pointer. Cannot be null.	Input
pstDst	Output image pointer. Cannot be null. Width and height same as pstSrc.	Output
pst16BitTo8BitCtrl	Control parameter pointer. Cannot be null.	Input
bInstant	Reserved.	Input

Parameter Name	Supported Image Type	Address Alignment	Resolution
pstSrc	U16C1, S16C1	2 byte	64x64 ~ 1920x1080
pstDst	U8C1, S8C1	1 byte	Same as pstSrc

Return Value

Return Value Description

0 Successful.

Non-zero Failed. Please refer to ERROR CODE.
Requirement
- Header: mi_common_datatype.h, mi_ive.h, mi_ive_datatype.h
- Library: libmi_ive.a
Note
- 4 modes are available for configuration. For details, please refer to MI_IVE_16BitTo8BitMode_e.
- The calculation formulas of different mode are as follows:
  - E_MI_IVE_16BIT_TO_8BIT_MODE_S16_TO_S8:
    
    $I_{out}(x,y)=\left\{ \begin{aligned} -128,(\frac{a}{b}I(x,y)<-128) \\ \frac{a}{b}I(x,y),(-128\leqslant \frac{a}{b}I(x,y)\leqslant 127) \\ 127,(\frac{a}{b}I(x,y)>127) \end{aligned} \right.$
    
    E_MI_IVE_16BIT_TO_8BIT_MODE_S16_TO_U8_ABS:
    
    $I_{out}(x,y)=\left\{ \begin{aligned} \left |\frac{a}{b}I(x,y) \right|,(\left| \frac{a}{b}I(x,y)\right|\leqslant 255) \\ 255,(\left| \frac{a}{b}I(x,y)\right|> 255) \end{aligned} \right.$
  - E_MI_IVE_16BIT_TO_8BIT_MODE_S16_TO_U8_BIAS:
    
    $I_{out}(x,y)=\left\{ \begin{aligned} 0,(\frac{a}{b}I(x,y)+bais< 0) \\ \frac{a}{b}I(x,y)+bais,(0\leqslant\frac{a}{b}I(x,y)+bais\leqslant 255) \\ 255, (\frac{a}{b}I(x,y)+bais> 255) \end{aligned} \right.$
  - E_MI_IVE_16BIT_TO_8BIT_MODE_U16_TO_U8:
    
    $I_{out}(x,y)=\left\{ \begin{aligned} 0,(\frac{a}{b}I(x,y)< 0) \\ \frac{a}{b}I(x,y),(0\leqslant\frac{a}{b}I(x,y)\leqslant 255) \\ 255, (\frac{a}{b}I(x,y)> 255) \end{aligned} \right.$
  Where,
  - $I(x,y)$ refers to pstSrc, $I_{out}(x,y)$ refers to pstDst.
  - $mode$ , $a$ , $b$ amd $bias$ refer respectively to eMode, u8Numerator, u16Denominator, and s8Bias in pst16BitTo8BitCtrl.
  - Requirement: $u8Numerator ≤ u16Denominator$ , and $u16Denominator\neq0$ .
Related API

MI_IVE_ThreshS16

MI_IVE_ThreshU16

2.20. MI_IVE_OrdStatFilter¶

Function

Execute a 3x3 template sequential statistic filtering task in which Median, Max, and Min filtering are supported.

Syntax

MI_S32 MI_IVE_OrdStatFilter(MI_IVE_HANDLE hHandle, MI_IVE_SrcImage_t *pstSrc, MI_IVE_DstImage_t *pstDst, MI_IVE_OrdStatFilter_t *pstOrdStatFltCtrl, MI_BOOL bInstant);

Parameter

Parameter Name	Description	Input/Output
hHandle	Regional handle number. Parameter range: [0, MI_IVE_HANDLE_MAX).	Input
pstSrc	Source image pointer. Cannot be null.	Input
pstDst	Output image pointer. Cannot be null. Width and height same as pstSrc.	Output
pstOrdStatFltCtrl	Control parameter pointerCannot be null.	Input
bInstant	Reserved.	Input

Parameter Name	Supported Image Type	Address Alignment	Resolution
pstSrc	U8C1	16 byte	64x64 ~ 1920x1080
pstDst	U8C1	16 byte	Same as pstSrc

For Pcupid series: The hardware acceleration is supported when the resolution is not exceeded 640x480.

Return Value

Return Value Description

0 Successful.

Non-zero Failed. Please refer to ERROR CODE.
Requirement
- Header: mi_common_datatype.h, mi_ive.h, mi_ive_datatype.h
- Library: libmi_ive.a
Note
- Three filter modes are available for configuration. For details, please refer to MI_IVE_OrdStatFilterMode_e.
- The calculation formula of different mode are as follows:
  - E_MI_IVE_ORD_STAT_FILTER_MODE_MEDIAN:
    
    $I_{out}(x,y)=median_{-1\leqslant i\leqslant1,-1\leqslant j\leqslant1}\{I(x+i,y+j)\}$
  - E_MI_IVE_ORD_STAT_FILTER_MODE_MAX:
    
    $I_{out}(x,y)=max_{-1\leqslant i\leqslant1,-1\leqslant j\leqslant1}\{I(x+i,y+j)\}$
  - E_MI_IVE_ORD_STAT_FILTER_MODE_MIN:
    
    $I_{out}(x,y)=min_{-1\leqslant i\leqslant1,-1\leqslant j\leqslant1}\{I(x+i,y+j)\}$
  Where, $I(x,y)$ refers to pstSrc and $I_{out}(x,y)$ refers to pstDst.
Related API

MI_IVE_Filter

MI_IVE_Dilate

MI_IVE_Erode

Example

MI_S32 Sample_OrdStatFilter()
{
    MI_IVE_HANDLE handle = 0;
    MI_IVE_SrcImage_t stSrc = {0};
    MI_IVE_DstImage_t stDst = {0};
    MI_IVE_OrdStatFilter_t stCtrl =
    {
        .eMode = E_MI_IVE_ORD_STAT_FILTER_MODE_MEDIAN, // for median filter
    };

    // create IVE handle
    if (MI_SUCCESS != (s32Ret = MI_IVE_Create(handle)))
    {
        printf("Could not create IVE handle\n");
        return E_MI_ERR_FAILED;
    }

    while (loop_condition == true)
    {
        // ...set input image info to stSrc. e.g. stSrc.eType = E_MI_IVE_IMAGE_TYPE_U8C1;
        ...
        // ...set output image info to stDst. e.g. stDst.eType = E_MI_IVE_IMAGE_TYPE_U8C1;
        ...

        // OrdStatFilter task.
        if (MI_SUCCESS != (s32Ret = MI_IVE_OrdStatFilter(handle, &stSrc, &stDst, &stCtrl, 0)))
        {
            printf("MI_IVE_OrdStatFilter() return ERROR 0x%X\n", s32Ret);
            MI_IVE_Destroy(handle);
            return E_MI_ERR_FAILED;
        }

        // get output from stDst and do something expected.
        ...
    }

    // destroy IVE handle
    MI_IVE_Destroy(handle);
    return MI_SUCCESS;
}

2.21. MI_IVE_Map¶

Function

Execute a Map (mapping assignment) task, by looking up the Map to look for the value for each pixel of the source image in the lookup table, and assigning to the target image the value in the corresponding pixel lookup table. U8C1 to U8C1 mode mapping is supported.

Syntax

MI_S32 MI_IVE_Map(MI_IVE_HANDLE hHandle, MI_IVE_SrcImage_t *pstSrc, MI_IVE_SrcMemInfo_t *pstMap, MI_IVE_DstImage_t *pstDst, MI_BOOL bInstant);

Parameter

Parameter Name	Description	Input/Output
hHandle	Regional handle number. Parameter range: [0, MI_IVE_HANDLE_MAX).	Input
pstSrc	Source image pointer. Cannot be null.	Input
pstMap	Mapping table info pointer. Cannot be null. The memory should at least have the size of (MI_IVE_MapLutMem_t).	Input
pstDst	Output image pointer. Cannot be null. Width and height same as pstSrc.	Output
bInstant	Reserved.	Input

Parameter Name	Supported Image Type	Address Alignment	Resolution
pstSrc	U8C1	1 byte	64x64 ~ 1920x1080
pstMap	-	16 byte	-
pstDst	U8C1	1 byte	Same as pstSrc

Return Value

Return Value Description

0 Successful.

Non-zero Failed. Please refer to ERROR CODE.
Requirement
- Header: mi_common_datatype.h, mi_ive.h, mi_ive_datatype.h
- Library: libmi_ive.a
Note
- The calculation formula of Map is as follows:
  
  $I_{out}(x,y)=map[I(xy)]$
  
  Where, $I(x,y)$ refers to pstSrc, $I_{out}(x,y)$ refers to pstDst, and $map$ refers to pstMap.

2.22. MI_IVE_EqualizeHist¶

Function

Execute a grayscale-image histogram equalization calculation task.

Syntax

MI_S32 MI_IVE_EqualizeHist(MI_IVE_HANDLE hHandle, MI_IVE_SrcImage_t *pstSrc, MI_IVE_DstImage_t *pstDst, MI_IVE_EqualizeHistCtrl_t *pstEqualizeHistCtrl, MI_BOOL bInstant);

Parameter

Parameter Name	Description	Input/Output
hHandle	Regional handle number. Parameter range: [0, MI_IVE_HANDLE_MAX).	Input
pstSrc	Source image pointer. Cannot be null.	Input
pstDst	Output image pointer. Cannot be null. Width and height same as pstSrc.	Output
pstEqualizeHistCtrl	Control parameter pointer. Cannot be null.	Input
bInstant	Reserved.	Input

Parameter Name	Supported Image Type	Address Alignment	Resolution
pstSrc	U8C1	16 byte	64x64 ~ 1920x1080
pstDst	U8C1	16 byte	Same as pstSrc
pstEqualizeHistCtrl→stMem	-	16 byte	-

Return Value

Return Value Description

0 Successful.

Non-zero Failed. Please refer to ERROR CODE.
Requirement
- Header: mi_common_datatype.h, mi_ive.h, mi_ive_datatype.h
- Library: libmi_ive.a
Note
- stMem in pstEqualizeHistCtrl should at least have the size of MI_IVE_EqualizeHistCtrlMem_t and agree with the histogram equalization calculation process.

2.23. MI_IVE_Add¶

Function

Execute a weighted addition calculation task against two grayscale images.

Syntax

MI_S32 MI_IVE_Add(MI_IVE_HANDLE hHandle, MI_IVE_SrcImage_t *pstSrc1, MI_IVE_SrcImage_t *pstSrc2, MI_IVE_DstImage_t *pstDst, MI_IVE_AddCtrl_t *pstAddCtrl, MI_BOOL bInstant);

Parameter

Parameter Name	Description	Input/Output
hHandle	Regional handle number. Parameter range: [0, MI_IVE_HANDLE_MAX).	Input
pstSrc1	Source image 1 pointer. Cannot be null.	Input
pstSrc2	Source image 2 pointer. Cannot be null. Width and height same as pstSrc1.	Input
pstDst	Output image pointer. Width and height same as pstSrc1; Cannot be null.	Output
pstAddCtrl	Control parameter pointer. Cannot be null.	Input
bInstant	Reserved.	Input

Parameter Name	Supported Image Type	Address Alignment	Resolution
pstSrc1	U8C1	1 byte	64x64 ~ 1920x1080
pstSrc2	U8C1	1 byte	Same as pstSrc
pstDst	U8C1	1 byte	Same as pstSrc

Return Value

Return Value Description

0 Successful.

Non-zero Failed. Please refer to ERROR CODE.
Requirement
- Header: mi_common_datatype.h, mi_ive.h, mi_ive_datatype.h
- Library: libmi_ive.a
Note
- The calculation formula of Add is as follows:
  
  $I_{out}(x,y)=x*I_{src1}(x,y)+y*I_{src2}(x,y)$
  
  Where,
  - $I_{src1}(i,j)$ refers to pstSrc1, $I_{src2}(i,j)$ refers to pstSrc2, $I_{out}(i,j)$ refers to pstDst.
  - $x$ , $y$ refer to u0q16X and u0q16Y in pstAddCtrl. It is required that $0<x<1, 0<y<1$ , and $\ x+y=1$ before the fixed point.
Related API

MI_IVE_Sub

2.24. MI_IVE_Xor¶

Function

Execute an XOR calculation task against two binary graphs.

Syntax

MI_S32 MI_IVE_Xor(MI_IVE_HANDLE hHandle, MI_IVE_SrcImage_t *pstSrc1, MI_IVE_SrcImage_t *pstSrc2, MI_IVE_DstImage_t *pstDst, MI_BOOL bInstant);

Parameter

Parameter Name	Description	Input/Output
hHandle	Regional handle number. Parameter range: [0, MI_IVE_HANDLE_MAX).	Input
pstSrc1	Source image 1 pointer. Cannot be null.	Input
pstSrc2	Source image 1 pointer. Cannot be null. Width and height same as pstSrc1.	Input
pstDst	Output image pointer. Cannot be null. Width and height same as pstSrc1.	Output
bInstant	Reserved.	Input

Parameter Name	Supported Image Type	Address Alignment	Resolution
pstSrc1	U8C1	1 byte	64x64 ~ 1920x1080
pstSrc2	U8C1	1 byte	Same as pstSrc1
pstDst	U8C1	1 byte	Same as pstSrc1

Return Value

Return Value Description

0 Successful.

Non-zero Failed. Please refer to ERROR CODE.
Requirement
- Header: mi_common_datatype.h, mi_ive.h, mi_ive_datatype.h
- Library: libmi_ive.a
Note
- The calculation formula of Xor is as follows:
  
  $I_{out}(x,y)=I_{src1}(x,y)\bigoplus{I_{src2}(x,y)}$
  
  Where, $I_{src1}(x,y)$ refers to pstSrc1, $I_{src2}(x,y)$ refers to pstSrc2, and $I_{out}(x,y)$ refers to pstDst.
Related API

MI_IVE_And

MI_IVE_Or

2.25. MI_IVE_Ncc¶

Function

Execute a normalized cross-correlation calculation task against two grayscale images of the same resolution.

Syntax

MI_S32 MI_IVE_Ncc(MI_IVE_HANDLE hHandle, MI_IVE_SrcImage_t *pstSrc1, MI_IVE_SrcImage_t *pstSrc2, MI_IVE_DstMemInfo_t *pstDst, MI_BOOL bInstant);

Return Value

Parameter Name	Description	Input/Output
hHandle	Regional handle number. Parameter range: [0, MI_IVE_HANDLE_MAX).	Input
pstSrc1	Source image 1 pointer. Cannot be null.	Input
pstSrc2	Source image 2 pointer. Cannot be null. Width and height same as pstSrc1.	Input
pstDst	Output data pointer. Cannot be null. The memory should at least have the size of (MI_IVE_NccDstMem_t).	Output
bInstant	Reserved.	Input

Parameter Name	Supported Image Type	Address Alignment	Resolution
pstSrc1	U8C1	1 byte	32x32 ~ 1920x1080
pstSrc2	U8C1	1 byte	Same as pstSrc1
pstDst	-	16 byte	-

Return Value

Return Value Description

0 Successful.

Non-zero Failed. Please refer to ERROR CODE.
Requirement
- Header: mi_common_datatype.h, mi_ive.h, mi_ive_datatype.h
- Library: libmi_ive.a
Note
- The calculation formula of NCC is as follows:
  
  $NCC(I_{src1},I_{src2})=\frac{\sum^w_{i-1}\sum^h_{j=1}(I_{src1}(i,j)*I_{src2}(i,j))}{\sqrt{\sum^w_{i=1}\sum^h_{j=1}(I^2_{src1}(i,j))}\sqrt{\sum^w_{i=1}\sum^h_{j=1}(I^2_{src2}(i,j))}}$
  
  Only output the numerator and the two denominators before square root in the above formula:
  - pstDst→u64Numerator= $\sum^w_{i-1}\sum^h_{j=1}(I_{src1}(i,j)*I_{src2}(i,j))$
  - pstDst→u64QuadSum1= $\sum^w_{i=1}\sum^h_{j=1}(I^2_{src1}(i,j))$
  - pstDst→u64QuadSum2= $\sum^w_{i=1}\sum^h_{j=1}(I^2_{src2}(i,j))$ .

2.26. MI_IVE_Ccl¶

Function

Execute a connected region label task against binary images.

Syntax

MI_S32 MI_IVE_Ccl(MI_IVE_HANDLE hHandle, MI_IVE_Image_t *pstSrcDst, MI_IVE_DstMemInfo_t *pstBlob, MI_IVE_CclCtrl_t *pstCclCtrl, MI_BOOL bInstant);

Parameter

Parameter Name	Description	Input/Output
hHandle	Regional handle number. Parameter range: [0, MI_IVE_HANDLE_MAX).	Input
pstSrcDst	Source image pointer. Connected region is labeled on the source image, i.e. source image is also the labeled image output. Cannot be null.	Input, Output
pstBlob	Connected region info pointer. Cannot be null. The memory should at least have the size of (MI_IVE_CcBlob_t), and output at most 254 valid connected regions.	Output
pstCclCtrl	Control parameter pointer. Cannot be null.	Input
bInstant	Reserved.	Input

Parameter Name	Supported Image Type	Address Alignment	Resolution
pstSrcDst	U8C1	16 byte	16x4~720x640
pstBlob	-	16 byte	-

For Souffle series: When the parameters meet the limitation of hardware acceleration usage, the maximum supported resolution is 1280x720.

Return Value

Return Value Description

0 Successful.

Non-zero Failed. Please refer to ERROR CODE.
Requirement
- Header: mi_common_datatype.h, mi_ive.h, mi_ive_datatype.h
- Library: libmi_ive.a
Note
- The labeled results are stored in the virtual address pstBlob→pu8VirAddr, use a pointer variable of MI_IVE_CcBlob_t to point to this address as below, the labeled results could be got by operating this pointer, for details, please refer to MI_IVE_CcBlob_t.
  
  MI_IVE_CcBlob_t* ccBlob = (MI_IVE_CcBlob_t*) pstBlob->pu8VirAddr;

2.27. MI_IVE_Gmm¶

Function

Execute a GMM background modelling task. Grayscale image is supported by RGB_PACKAGE image GMM background modelling. The Gaussian model number is 3 or 5.

Syntax

MI_S32 MI_IVE_Gmm(MI_IVE_HANDLE hHandle, MI_IVE_SrcImage_t *pstSrc, MI_IVE_DstImage_t *pstFg, MI_IVE_DstImage_t *pstBg, MI_IVE_MemInfo_t *pstModel, MI_IVE_GmmCtrl_t *pstGmmCtrl, MI_BOOL bInstant);

Parameter

Parameter Name	Description	Input/Output
hHandle	Regional handle number. Parameter range: [0, MI_IVE_HANDLE_MAX).	Input
pstSrc	Source image pointer. Cannot be null.	Input
pstFg	Foreground image pointer. Cannot be null. Width and height same as pstSrc.	Output
pstBg	Background image pointer. Cannot be null. Width and height same as pstSrc.	Output
pstModel	GMM modelling parameter pointer. Cannot be null.	Input, Output
pstGmmCtrl	Control parameter pointer. Cannot be null.	Input
bInstant	Reserved.	Input

Parameter Name	Supported Image Type	Address Alignment	Resolution
pstSrc	U8C1, U8C3_PACKAGE	16 byte	64x64~1280x720
pstFg	U8C1 binary image	16 byte	Same as pstSrc
pstBg	Same as pstSrc	16 byte	Same as pstSrc
pstModel	-	16 byte	-

Return Value

Return Value Description

0 Successful.

Non-zero Failed. Please refer to ERROR CODE.
Requirement
- Header: mi_common_datatype.h, mi_ive.h, mi_ive_datatype.h
- Library: libmi_ive.a

Note

The GMM implementation refer to MOG and MOG2 in OpenCV.

Grayscale image GMM employs n (n=3 or 5) Gaussian models. The memory allocation of model data is as illustrated below.

Fig 1-10: Grayscale Image GMM Model Memory Allocation

Chips/Platforms	Model_cnt	Weight	Mean	Var	Aligment Requirement	Required Memory Size m of 3 Model	Required Memory Size m of 5 Model
Souffle	3 bits	2 bytes	2 bytes	3 bytes	16-bytes-aligned	3 bits + 3*(2+2+3) bytes m is 32 bytes after alignment	3 bits + 5*(2+2+3) bytes m is 48 bytes after alignment
Others	-	2 bytes	2 bytes	3 bytes	-	3*(2+2+3) bytes m is 21 bytes	5*(2+2+3) bytes m is 35 bytes

RGB image GMM employs n (n=3 or 5) Gaussian models. The memory allocation of pstModel is as illustrated below.

Fig 1-11: RGB Image GMM Model Memory Allocation

Chips/Platforms	Model_cnt	Weight	Mean	Var	Aligment Requirement	Required Memory Size m of 3 Model	Required Memory Size m of 5 Model
Souffle	3 bits	2 bytes	3*2 bytes	3 bytes	16-bytes-aligned	3 bits + 3*(2+6+3) bytes m is 48 bytes after alignment	3 bits + 5*(2+6+3) bytes m is 64 bytes after alignment
Others	-	4 bytes	3*4 bytes	4 bytes	-	3*(4+12+4) bytes m is 60 bytes	5*(4+12+4) bytes m is 100 bytes

According to the calculated $m$ in above tabulation, the required memory size of model with different model num and output format can be represented as follows:

pstModel→u32Size = $m$ * pstSrc→u16Width * pstSrc→u16Height

2.28. MI_IVE_CannyHysEdge¶

Function

Execute the first half of Canny edge extraction task against grayscale images: Gradient, gradient amplitude calculation, hysteresis thresholding and non-maximum suppression.

Syntax

MI_S32 MI_IVE_CannyHysEdge(MI_IVE_HANDLE hHandle, MI_IVE_SrcImage_t *pstSrc, MI_IVE_DstImage_t *pstEdge, MI_IVE_DstMemInfo_t *pstStack, MI_IVE_CannyHysEdgeCtrl_t *pstCannyHysEdgeCtrl, MI_BOOL bInstant);

Parameter

Parameter Name	Description	Input/Output
hHandle	Regional handle number. Parameter range: [0, MI_IVE_HANDLE_MAX).	Input
pstSrc	Source image pointer. Cannot be null.	Input
pstEdge	Strong and weak edge mark image pointer. Cannot be null. Width and height same as pstSrc.	Output
pstStack	Strong edge point coordinate stack. Cannot be null. Memory should at least be: pstSrc→u16Width * pstSrc→u16Height * (sizeof(MI_IVE_PointU16_t)) + sizeof(MI_IVE_CannyStackSize_t)	Output
pstCannyHysEdgeCtrl	Control parameter pointer. Cannot be null.	Input
bInstant	Reserved.	Input

Parameter Name	Supported Image Type	Address Alignment	Resolution
pstSrc	U8C1	16 byte	64x64 ~ 1920x1080
pstEdge	U8C1	16 byte	Same as pstSrc
pstStack	-	16 byte	-
pstCannyHysEdgeCtrl→stMem	-	16 byte	-

Return Value

Return Value Description

0 Successful.

Non-zero Failed. Please refer to ERROR CODE.
Requirement
- Header: mi_common_datatype.h, mi_ive.h, mi_ive_datatype.h
- Library: libmi_ive.a
Note
- pstEdge has only three values: 0, 1, and 2:
  
  0 means weak edge point.
  
  1 means non-edge point
  
  2 means strong edge point.
- pstStack stores the coordinate information of strong edge point.
- pstCannyHysEdgeCtrl→stMem requires at least the following memory size: pstCannyHysEdgeCtrl→stMem.u32Size = pstSrc→azu16Stride[0] * (pstSrc→u16Height + 3) * 4.
- After completing the task, you must call the MI_IVE_CannyEdge function to output Canny edge image.
Related API

MI_IVE_CannyEdge

2.29. MI_IVE_CannyEdge¶

Function

Execute the second half of Canny edge extraction task against grayscale images: connecting edge points to form a Canny edge map.

Syntax

MI_S32 MI_IVE_CannyEdge(MI_IVE_HANDLE hHandle, MI_IVE_Image_t *pstEdge, MI_IVE_MemInfo_t *pstStack, MI_BOOL bInstant);

Parameter

Parameter Name	Description	Input/Output
hHandle	Regional handle number. Parameter range: [0, MI_IVE_HANDLE_MAX).	Input
pstEdge	Strong and weak edge mark image pointer when used as an input; edge binary image pointer when used as an output. Cannot be null.	Input, Output
pstStack	Strong edge point coordinate stack. Cannot be null.	Input, Output
bInstant	Reserved.	Input

Parameter Name	Supported Image Type	Address Alignment	Resolution
pstEdge	U8C1	16 byte	64x64 ~ 1920x1080
pstStack	-	16 byte	-

Return Value

Return Value Description

0 Successful.

Non-zero Failed. Please refer to ERROR CODE.
Requirement
- Header: mi_common_datatype.h, mi_ive.h, mi_ive_datatype.h
- Library: libmi_ive.a
Note
- Before using the interface, you must call MI_IVE_CannyHysEdge first. When the MI_IVE_CannyHysEdge task is finished, you can use the MI_IVE_CannyHysEdge pstEdge, pstStack output as the parameter input of the interface.
Related API

MI_IVE_CannyHysEdge

2.30. MI_IVE_Lbp¶

Function

Execute an LBP calculation task.

Syntax

MI_S32 MI_IVE_Lbp(MI_IVE_HANDLE hHandle, MI_IVE_SrcImage_t *pstSrc1, MI_IVE_SrcImage_t *pstSrc2, MI_IVE_DstImage_t *pstDst, MI_IVE_LbpCtrrl_t *pstLbpCtrl, MI_BOOL bInstant);

Parameter

Parameter Name	Description	Input/Output
hHandle	Regional handle number. Parameter range: [0, MI_IVE_HANDLE_MAX).	Input
pstSrc1	Source image pointer. Cannot be null.	Input
pstSrc2	Source image pointer. Cannot be null. If U8C1 is the input channel mode, it can be null.	Input
pstDst	Output image pointer. Cannot be null. Width and height same as pstSrc.	Output
pstLbpCtrl	Control info pointer. Cannot be null.	Input
bInstant	Reserved.	Input

Parameter Name	Supported Image Type	Address Alignment	Resolution
pstSrc	U8C1	16 byte	64x64 ~ 1920x1080
pstDst	U8C1	16 byte	Same as pstSrc

Return Value

Return Value Description

0 Successful.

Non-zero Failed. Please refer to ERROR CODE.
Requirement
- Header: mi_common_datatype.h, mi_ive.h, mi_ive_datatype.h
- Library: libmi_ive.a
Note
- The LBP calculation formula of U8C1/U8C2 mode is as illustrated in the following figure.
  - U8C1 mode:
  - U8C2 mode:
- The calculation formulas of different mode are as follows:
  - E_MI_IVE_LBP_CMP_NORMAL
    
    $lbp(x,y)=\sum_{i=0}^7(I_i-I_c) \geq thr) << (7-i), \ thr \in [-128, 127]$
  - E_MI_IVE_LBP_CMP_ABS
    
    $lbp(x,y)=\sum_{i=0}^7(abs(I_i-I_c) \geq thr) << (7-i), \ thr \in [0, 255]$
  - E_MI_IVE_LBP_CMP_ABS_MUL
    
    $lbp(x,y)=\sum_{i=0}^7((I_i-I_c) \geq thr*I_c) << (7-i), \ thr \in [0, 1]$
  Where,
  - $I_i$ refers to pstSrc1, $lpb(x,y)$ refers to pstDst, and $thr$ refers to pstLbpCtrl→un8BitThr.
  - In U8C1 mode, $I_c$ refers to pstScr1, In U8C2 mode, $I_c$ refers to pstScr2.

2.31. MI_IVE_NormGrad¶

Function

Execute a normalized gradient calculation task, in which the gradient average components are normalized to S8.

Syntax

MI_S32 MI_IVE_NormGrad(MI_IVE_HANDLE hHandle, MI_IVE_SrcImage_t *pstSrc, MI_IVE_DstImage_t *pstDstH, MI_IVE_DstImage_t *pstDstV, MI_IVE_DstImage_t *pstDstHV, MI_IVE_NormGradCtrl_t *pstNormGradCtrl, MI_BOOL bInstant);

Parameter

Parameter Name	Description	Input/Output
hHandle	Regional handle number. Parameter range: [0, MI_IVE_HANDLE_MAX).	Input
pstSrc	Source image pointer. Cannot be null.	Input
pstDstH	Gradient component image H pointer gained by template filtering and normalization to S8. According to pstNormGradCtrl→eOutCtrl, this parameter cannot be null if
output is required.	Output
pstDstV	Gradient component image V pointer gained by transposed template filtering and normalization to S8. According to pstNormGradCtrl→eOutCtrl, this parameter cannot be null if
output is required.	Output
pstDstHV	Image pointer gained by template and transposed template filtering and normalization to S8. The image pointer is saved in package format. According to pstNormGradCtrl→eOutCtrl, this parameter cannot be null if
output is required.	Output
pstNormGradCtrl	Control info pointer.	Input
bInstant	Reserved.	Input

Parameter Name	Supported Image Type	Address Alignment	Resolution
pstSrc	U8C1	16 byte	64x64 ~ 1920x1080
pstDstH	S8C1	16 byte	Same as pstSrc
pstDstV	S8C1	16 byte	Same as pstSrc
pstDstHV	S8C2_PACKAGE	16 byte	Same as pstSrc

Return Value

Return Value Description

0 Successful.

Non-zero Failed. Please refer to ERROR CODE.
Requirement
- Header: mi_common_datatype.h, mi_ive.h, mi_ive_datatype.h
- Library: libmi_ive.a
Note
- When the output mode is E_MI_IVE_NORM_GRAD_OUT_CTRL_HOR_AND_VER, pstDstH and pstDstV pointers cannot be null and the stride of pstDstH and pstDstV must be consistent.
- Four output formats are available for configuration. For details, please refer to MI_IVE_NormGradOutCtrl_e.
- The calculation formula of NormGrad is as follows:
  
  $H_{out}={\sum_{-2<j<2}\sum_{-2<i<2}I(x+i,y+j)*coef(x+i,y+j)} >> norm$
  
  $V_{out}={\sum_{-2<j<2}\sum_{-2<i<2}I(x+i,y+j)*coef(x+j,y+i)} >> norm$
Related API

MI_IVE_Sobel

2.32. MI_IVE_LkOpticalFlow¶

Function

Execute a single layer LK optical flow calculation task.

Syntax

MI_S32 MI_IVE_LKOpticalFlow(MI_IVE_HANDLE hHandle, MI_IVE_SrcImage_t*pstSrcPre, MI_IVE_SrcImage_t*pstSrcCur, MI_IVE_SrcMemInfo_t *pstPoint, MI_IVE_MemInfo_t *pstMv, MI_IVE_LkOpticalFlowCtrl_t *pstLkOptiFlowCtrl, MI_BOOL bInstant);

Parameter

Parameter Name	Description	Input/Output
hHandle	Regional handle number. Parameter range: [0, MI_IVE_HANDLE_MAX).	Input
pstSrcPre	Previous image pointer. Cannot be null.	Input
pstSrcCur	Current image pointer. Cannot be null. Width and height same as pstSrcPre.	Input
pstPoint	Current pyramid layer initial feature point coordinate. Cannot be null. The coordinate can only be of the type MI_IVE_PointS25Q7_t. The memory should at least have the size below: pstLkOptiFlowCtrl→u16CornerNum * sizeof(MI_IVE_PointS25Q7_t).	Input
pstMv	Corresponds to pstPoint feature point motion displacement vector. Cannot be null. The first calculation needs to be initialized to 0 input; the subsequent layer calculation should input the motion displacement vector calculated by the upper layer. The displacement can only be of the type MI_IVE_MvS9Q7_t. The memory should at least have the size below: pstLkOptiFlowCtrl→u16CornerNum * sizeof(MI_IVE_MvS9Q7_t)	Input, Output
pstLkOptiFlowCtrl	Control parameter pointer. Cannot be null.	Input
bInstant	Reserved.	Input

Parameter Name	Supported Image Type	Address Alignment	Resolution
pstSrcPre	U8C1	16 byte	64x64 ~ 1920x1080
pstSrcCur	U8C1	16 byte	Same as pstSrcPre

Return Value

Return Value Description

0 Successful.

Non-zero Failed. Please refer to ERROR CODE.
Requirement
- Header: mi_common_datatype.h, mi_ive.h, mi_ive_datatype.h
- Library: libmi_ive.a
Note
- In solving the following optical flow equation, only 7x7 pixels around the feature point are used to calculate the corresponding $I_x$ , $I_y$ , $I_t$ .
  
  $\begin{bmatrix} \sum I_{x}^{2} & \sum I_{x}I_{y} \\ \sum I_{x}I_{y} & \sum I_{y}^{2} \\ \end{bmatrix} \begin{bmatrix} u \\ v \end{bmatrix} \begin{bmatrix} -\sum I_{x}I_{t}\\ -\sum I_{y}I_{t} \\ \end{bmatrix}$
  
  Where, $I_x$ , $I_y$ , $I_t$ refer respectively to the partial derivation of the current image in the x, y direction and the difference between the current image and the previous image.
- Let’s take the following 3-layer pyramid LK optical flow calculation as an example. The width and height of the image of each layer must be half the width and height of the image of its upper layer. The calculation method is as shown in the following figure.
  
  Fig 1-12: 3-Layer Pyramid LK Optical Flow Calculation
  - According to the input feature point coordinates, the coordinates corresponding to the feature points of the three-layer pyramid are calculated: p0, p1, p2;
  - Using p2 and mv2, which is initially 0, as input, call the LK operator to find the displacement mv2 on the second layer;
  - Using p1 and mv2 as input, call the LK operator to find the displacement mv1 on the frist layer;
  - Using p0 and mv1 as input, call the LK operator to find the displacement mv0 on the zero layer;
  - If the zero layer is not the original image, the true displacement mv of the LK optical flow can be obtained according to the proportional relationship between the zero layer and the original image.
- Design and usage restrictions: Each feature point is calculated only by the data of the fixed size window centered on the feature point. If the feature point displacement target point exceeds the fixed size window during the iterative calculation process, the calculation optical flow would fail.

2.33. MI_IVE_Sad¶

Function

Calculate 16-bit/8-bit SAD images of 4x4/8x8/16x16 blocks for two images, and threshold output for SAD

Syntax

MI_S32 MI_IVE_Sad(MI_IVE_HANDLE hHandle, MI_IVE_SrcImage_t*pstSrc1, MI_IVE_SrcImage_t*pstSrc2, MI_IVE_DstImage_t *pstSad, MI_IVE_DstImage_t *pstThr, MI_IVE_SadCtrl_t *pstSadCtrl, MI_BOOL bInstant);

Parameter

Parameter Name	Description	Input/Output
hHandle	Regional handle number. Parameter range: [0, MI_IVE_HANDLE_MAX).	Input
pstSrc1	Source image 1 pointer. Cannot be null.	Input
pstSrc2	Source image 2 pointer. Cannot be null. Width and height same as pstSrc1.	Input
pstSad	Output SAD image pointer.	Output
pstThr	Output SAD thresholding image pointer. According to pstSadCtrl→eOutCtrl, this parameter cannot be null if output is required. According to pstSadCtrl→eMode, refers to 4x4, 8x8, and 16x16 block modes, the width and height thereof being ¼, ⅛, and 1/16 of pstSrc1 respectively. According to pstSadCtrl→eOutCtrl, this parameter cannot be null if output is required. According to pstSadCtrl→eMode, refers to 4x4, 8x8, and 16x16 block modes, the width and height thereof being ¼, ⅛, and 1/16 of pstSrc1 respectively.	Output
pstSadCtrl	Control info pointer. Cannot be null.	Input
bInstant	Reserved.	Input

Parameter Name	Supported Image Type	Address Alignment	Resolution
pstSrc1	U8C1	1 byte	64x64 ~ 1920x1080
pstSrc2	U8C1	1 byte	Same as pstSrc1
pstSad	U8C1, U16C1	16 byte	According to pstSadCtrl→eMode, refers to 4x4, 8x8, and 16x16 block modes, the width and height thereof being ¼, ⅛, and 1/16 of pstSrc1 respectively.
pstThr	U8C1	16 byte	According to pstSadCtrl→eMode, refers to 4x4, 8x8, 16x16 block modes, the width and height thereof being ¼, ⅛, and 1/16 of pstSrc1 respectively.

Return Value

Return Value Description

0 Successful.

Non-zero Failed. Please refer to ERROR CODE.
Requirement
- Header: mi_common_datatype.h, mi_ive.h, mi_ive_datatype.h
- Library: libmi_ive.a
Note
- The calculation formula of SAD is as follows:
  
  $SAD_{out}(x,y)=\sum_{i=0}^{n-1}\sum_{j=0}^{n-1}abs(I_{src1}(n*x+i, n*y+j) - I_{src2}(n*x+i, n*y+j))$
  
  $Thr(x,y)=\begin{cases} minVal \quad O(x,y) \leq Thresh \\ maxVal \quad O(x,y) > Thresh \\ \end{cases}$
  
  Where,
  - $I_{src1}(i,j)$ refers to pstSrc1, $I_{src2}(i,j)$ refers to pstSrc2, $SAD_{out}(x,y)$ refers to pstSad, $Thr(x,y)$ refers to pstThr;
  - $Thresh$ , $minVal$ and $maxVal$ refer respectively to pstSadCtrl→u16Thr, pstSadCtrl→u8MinVal and stSadCtrl→u8MaxVal.
  - $n$ is related to pstSadCtrl→eMode and represents 4, 8 or 16 respectively when the eMode is E_MI_IVE_SAD_MODE_MB_4X4, E_MI_IVE_SAD_MODE_MB_8X8 or E_MI_IVE_SAD_MODE_MB_16X16;

2.34. MI_IVE_Bernsen¶

Function

Execute a Bernsen thresholding task for the 3x3 and 5x5 windows.

Syntax

MI_S32 MI_IVE_Bernsen(MI_IVE_HANDLE hHandle, MI_IVE_SrcImage_t *pstSrc, MI_IVE_DstImage_t *pstDst, MVE_IVE_BernsenCtrl_t *pstBernsenCtrl, MI_BOOL bInstant);

Parameter

Parameter Name	Description	Input/Output
hHandle	Regional handle number. Parameter range: [0, MI_IVE_HANDLE_MAX).	Input
pstSrc	Source image pointer. Cannot be null.	Input
pstDst	Pointer to the output image template. It cannot be null. The height and width are the same as those of pstSrc.	Output
pstBernsenCtrl	Pointer to the control parameter. It cannot be null.	Output
bInstant	Reserved.	Input

Parameter Name	Supported Image Type	Address Alignment	Resolution
pstSrc	U8C1	16 byte	64x64 ~ 1920x1080
pstDst	S8C1	16 byte	Same as pstSrc

For Pcupid series: The hardware acceleration is supported when the resolution is not exceeded 640x480.

Return Value

Return Value Description

0 Successful.

Non-zero Failed. Please refer to ERROR CODE.
Requirement
- Header: mi_common_datatype.h, mi_ive.h, mi_ive_datatype.h
- Library: libmi_ive.a
Note
- Two modes are supported. See MI_IVE_BernsenMode_e for details.
- The calculation formulas of different mode are as follows:
  - MI_BERNSEN_MODE_NORMAL
    
    $T(x,y)=0.5*(max_{-\omega\leqslant i\leqslant \omega,-\omega\leqslant j\leqslant \omega}I(x+i,y+j)+min_{-\omega\leqslant i\leqslant \omega,-\omega\leqslant j\leqslant \omega}I(x+i,y+j))$
    
    $I_{out}(x,y)=\left \{ \begin{aligned} 0,I(x,y)< T(x,y) \\ 1,I(x,y)\geqslant T(x,y) \end{aligned} \right.$
    
    No value needs to be assigned to pstBernsenCtrl->u8Thr.
  - MI_BERNSEN_MODE_THRESH
    
    $T(x,y)=0.5*(max_{-\omega\leqslant i\leqslant \omega,-\omega\leqslant j\leqslant \omega}I(x+i,y+j)+min_{-\omega\leqslant i\leqslant \omega,-\omega\leqslant j\leqslant \omega}I(x+i,y+j))$
    
    $I_{out}(x,y)=\left \{ \begin{aligned} 0,I(x,y)< 0.5*(T(x,y)+Thr) \\ 1,I(x,y)\geqslant 0.5*(T(x,y)+Thr) \end{aligned} \right.$
  Where $I(x,y)$ refer to pstSrc, $I_{out}(x,y)$ refer to pstDst, and $Thr$ refer to u8thr in pstBernsenCtrl.

2.35. MI_IVE_LineFilterHor¶

Function

Execute a horizontal density filter task for binary images.

Syntax

MI_S32 MI_IVE_LineFilterHor(MI_IVE_HANDLE hHandle, MI_IVE_SrcImage_t *pstSrcDst, MVE_IVE_LineFilterHorCtrl_t *pstLineFilterHorCtrl, MI_BOOL bInstant);

Parameter

Parameter Name	Description	Input/Output
hHandle	Regional handle number. Parameter range: [0, MI_IVE_HANDLE_MAX).	Input
pstSrcDst	Pointer to the source image or the output after processing. It cannot be null.	Input / Output
pstLineFilterHorCtrl	Pointer to the control parameter. It cannot be null.	Input
bInstant	Reserved.	Input

Parameter Name	Supported Image Type	Address Alignment	Resolution
pstSrcDst	U8C1 binary image	16 byte	64x64 ~ 1920x1080

Return Value

Return Value Description

0 Successful.

Non-zero Failed. Please refer to ERROR CODE.
Requirement
- Header: mi_common_datatype.h, mi_ive.h, mi_ive_datatype.h
- Library: libmi_ive.a
Note
- The images are counted in the horizontal direction. Each line consists of line segments that appear alternatively (white line segments) and gaps between the line segments (black line segments). The length of a black line segment is satisfied the following condition, the block line segment is set to a white line segment.
- The calculation formula of LineFilterHor is as follows:
  1. Horizontally scan the binary image, and record the result, as shown in the figure below.
  2. Assume the current gap is line_black56, if conform to the condition as below, the black line segment will be set to a white line segment.
    
    Condition 1 : line_black34 > thr1
    
    Condition 2 : line_black78 > thr1
    
    Condition 3 :
    
    (line_white45 + line_black56 + line_white67) <= (line_white45 + line_white67)thr2
    
    Condition 4 : (line_white45 + line_white67) > 1
  3. Following these step (step 1 and step 2), until handle the whole binary image. The result is shown in figure below.
  4. Horizontally scan the binary image again. According to the relationship between the white line segment and black line segment, transform the black line segment into the white line segment.
    
    Assume the current gap is line_black34, if conform to the condition as below, the black line segment will be set to a white line segment.
    
    Condition 1 : line_block34 < thr3
    
    Condition 2 : line_white47 > 2thr3
    
    Condition 3 : line_white23 > 9
    
    Condition 4 : line_white23 < 3thr3
  5. Following step 3 and step 4, until handle the whole binary image.

2.36. MI_IVE_LineFilterVer¶

Function

Execute a vertical density filter task for binary images.

Syntax

MI_S32 MI_IVE_LineFilterVer(MI_IVE_HANDLE hHandle, MI_IVE_SrcImage_t *pstSrcDst, MVE_IVE_LineFilterVerCtrl_t *pstLineFilterVerCtrl, MI_BOOL bInstant);

Parameter

Parameter Name	Description	Input/Output
hHandle	Regional handle number. Parameter range: [0, MI_IVE_HANDLE_MAX).	Input
pstSrcDst	Pointer to the source image or the output after processing. It cannot be null.	Input / Output
pstLineFilterVerCtrl	Pointer to the control parameter. It cannot be null.	Input
bInstant	Reserved.	Input

Parameter Name	Supported Image Type	Address Alignment	Resolution
pstSrcDst	U8C1 binary image	16 byte	64x64 ~ 1920x1080

Return Value

Return Value Description

0 Successful.

Non-zero Failed. Please refer to ERROR CODE.
Requirement
- Header: mi_common_datatype.h, mi_ive.h, mi_ive_datatype.h
- Library: libmi_ive.a
Note
- The images are counted in the vertical direction. Each line consists of line segments that appear alternatively white line segments and gaps between the line segments (black line segments). The length of a black line segment is satisfied the following condition, the block line segment is set to a white line segment.
- The calculation formula of LineFilterVer is as follows:
  1. Vertically scan the binary image, and record the result, as shown in the figure below.
  2. Assume the current gap is line_black56, if conform to the condition as below, the black line segment will be set to a white line segment.
    
    Condition 1 : line_black56 < thr
    
    Condition 2 : line_black67 > 6 & line_whitr67 < 25
    
    Condition 3 : line_white45 < 12
  3. Following these step (step 1 and step 2), until handle the whole binary image.

2.37. MI_IVE_NoiseRemoveHor¶

Function

Execute a horizontal noise removal task for binary images.

Syntax

MI_S32 MI_IVE_NoiseRemoveHor(MI_IVE_HANDLE hHandle, MI_IVE_SrcImage_t *pstSrcDst, MVE_IVE_NoiseRemoveHorCtrl_t *pstNoiseRemoveHorCtrl, MI_BOOL bInstant);

Parameter

Parameter Name	Description	Input/Output
hHandle	Regional handle number. Parameter range: [0, MI_IVE_HANDLE_MAX).	Input
pstSrcDst	Pointer to the source image or the output after processing. It cannot be null.	Input / Output
pstNoiseRemoveHorCtrl	Pointer to the control parameter. It cannot be null.	Input
bInstant	Reserved.	Input

Parameter Name	Supported Image Type	Address Alignment	Resolution
pstSrcDst	U8C1 binary image	16 byte	64x64 ~ 1920x1080

Return Value

Return Value Description

0 Successful.

Non-zero Failed. Please refer to ERROR CODE.
Requirement
- Header: mi_common_datatype.h, mi_ive.h, mi_ive_datatype.h
- Library: libmi_ive.a
Note
- The images are counted in the horizontal direction. Each line consists of line segments that appear alternatively (white line segments) and gaps between the line segments (black line segments). If the length of a line segment is smaller than the thr1 or bigger than the thr2, convert the line segment into a gap.

2.38. MI_IVE_NoiseRemoveVer¶

Function

Execute a vertical noise removal task for binary images.

Syntax

MI_S32 MI_IVE_NoiseRemoveHor(MI_IVE_HANDLE hHandle, MI_IVE_SrcImage_t *pstSrcDst, MVE_IVE_NoiseRemoveVerCtrl_t *pstNoiseRemoveVerCtrl, MI_BOOL bInstant);

Parameter

Parameter Name	Description	Input/Output
hHandle	Regional handle number. Parameter range: [0, MI_IVE_HANDLE_MAX).	Input
pstSrcDst	Pointer to the source image or the output after processing. It cannot be null.	Input / Output
pstNoiseRemoveVerCtrl	Pointer to the control parameter. It cannot be null.	Input
bInstant	Reserved.	Input

Parameter Name	Supported Image Type	Address Alignment	Resolution
pstSrcDst	U8C1 binary image	16 byte	64x64 ~ 1920x1080

Return Value

Return Value Description

0 Successful.

Non-zero Failed. Please refer to ERROR CODE.
Requirement
- Header: mi_common_datatype.h, mi_ive.h, mi_ive_datatype.h
- Library: libmi_ive.a
Note
- The images are counted in the vertical direction. Each line consists of line segments that appear alternatively (white line segments) and gaps between the line segments (black line segments). If the length of a line segment is smaller than the thr1 or bigger than the thr2, convert the line segment into a gap.

2.39. MI_IVE_AdpThresh¶

Function

Execute an adaptive thresholding task.

Syntax

MI_S32 MI_IVE_AdpThresh(MI_IVE_HANDLE hHandle, MI_IVE_SrcImage_t *pstSrc, MI_IVE_SrcImage_t *pstInteg, MI_IVE_DstImage_t *pstDst, MVE_IVE_AdpThreshCtrl_t *pstAdpThrCtrl, MI_BOOL bInstant);

Parameter

Parameter Name	Description	Input/Output
hHandle	Regional handle number. Parameter range: [0, MI_IVE_HANDLE_MAX).	Input
pstSrc	Pointer to the source image or the output after processing. It cannot be null.	Input
pstInteg	Pointer to the integral image of the source image. It cannot be null.	Input
pstDst	Pointer to the target binary image. It cannot be null.	Output
pstAdpThrCtrl	Pointer to the control parameter. It cannot be null.	Input
bInstant	Reserved.	Input

Parameter Name	Supported Image Type	Address Alignment	Resolution
pstSrc	U8C1	16 byte	64x64 ~ 1920x1080
pstInteg	U32C1	16 byte	Same as pstSrc
pstDst	U8C1	16 byte	Same as pstSrc

Return Value

Return Value Description

0 Successful.

Non-zero Failed. Please refer to ERROR CODE.
Requirement
- Header: mi_common_datatype.h, mi_ive.h, mi_ive_datatype.h
- Library: libmi_ive.a
Note
- The calculation formula of AdpThresh is as follows:
  
  $T(x,y)=\frac{1}{w*h}*\sum_{i=-h/2}^{h/2}\sum_{j=-w/2}^{w/2}I(x+i,y+j)$
  
  $I_{out}(x,y)=\begin{cases} 255 \quad (I(x,y) > (T(x,y) * RateThr) - Offset) \quad |\quad (I(x,y) \geq ValueThr) \\ 0 \quad else \\ \end{cases}$
  
  $Integ(x,y)=\sum_{i=-h/2}^{h/2}\sum_{j=-w/2}^{w/2}I(x+i,y+j)$
  
  Where,
  - $w=u8HalfMaskx, h=u8HalfMasky$ , $RateThr=u8RateThr / 10$ , $Offset=s16Offset$ , $ValueThr=u8ValueThr$
  - $I(x, y)$ refer to pstSrc, $I_{out}(x,y)$ refer to pstDst and $Integ(x,y)$ refer to pstInteg.

2.40. MI_IVE_Resize¶

Function

Execute an image scaling task, the input image can be scaling down or up to the indicated size and as output image.

Syntax

MI_S32 MI_IVE_Resize(MI_IVE_HANDLE hHandle, MI_IVE_SrcImage_t *pstSrc, MI_IVE_DstImage_t *pstDst, MVE_IVE_ResizeCtrl_t * pstResizeCtrl, MI_BOOL bInstant);

Parameter

Parameter Name	Description	Input/Output
hHandle	Regional handle number. Parameter range: [0, MI_IVE_HANDLE_MAX).	Input
pstSrc	Pointer to the source image or the output after processing. It cannot be null.	Input
pstDst	Pointer to the target binary image. It cannot be null.	Output
pstResizeCtrl	Pointer to the control parameter. It cannot be null.	Input
bInstant	Reserved.	Input

Parameter Name	Supported Image Type	Address Alignment	Resolution
pstSrc	U8C1, U8C3_PLANAR, U8C3_PACKAGE and YUV420SP	16 byte	16x4 ~ 1920x1080
pstDst	U8C1, U8C3_PLANAR, U8C3_PACKAGE and YUV420SP	16 byte	Same as pstSrc

For Pcupid series: The hardware acceleration is supported when the format is U8C1 and the resolution is not exceeded 640x480.

Return Value

Return Value Description

0 Successful.

Non-zero Failed. Please refer to ERROR CODE.
Requirement
- Header: mi_common_datatype.h, mi_ive.h, mi_ive_datatype.h
- Library: libmi_ive.a
Note
- Resize is implemented by referring to the standard bilinear and area method.
  - For Souffle/Pcupid series: The hardware acceleration is supported when the scaling factor is not exceeded 16, e.g. 16x16 image can be scaled up to 256x256 image at the most, 512x256 image can be scaled down to 32x16 image at the most.

Example

MI_S32 Sample_Resize()
{
    MI_IVE_HANDLE handle = 0;
    MI_IVE_SrcImage_t stSrc = {0};
    MI_IVE_DstImage_t stDst = {0};
    MVE_IVE_ResizeCtrl_t stCtrl =
    {
        .enMode = E_MI_IVE_RESIZE_TYPE_U8C1,
        .eResizeMethod = E_MI_IVE_RESIZE_METHOD_BILINEAR,
        // .eResizeMethod = E_MI_IVE_RESIZE_METHOD_AREA,
    };

    // create IVE handle
    if (MI_SUCCESS != (s32Ret = MI_IVE_Create(handle)))
    {
        printf("Could not create IVE handle\n");
        return E_MI_ERR_FAILED;
    }

    while (loop_condition == true)
    {
        // ...set input image info to stSrc. e.g. stSrc.eType = E_MI_IVE_IMAGE_TYPE_U8C1;
        ...
        // ...set output image info to stDst. e.g. stDst.eType = E_MI_IVE_IMAGE_TYPE_U8C1;
        ...
        stDst.u16Width = TargetWidth; // output width after resize, and should be 16-pixel-align
        stDst.u16Height = TargetHeight // output height after resize, and should be 2-pixel-align

        // Resize task.
        if (MI_SUCCESS != (s32Ret = MI_IVE_Resize(handle, &stSrc, &stDst, &stCtrl, 0)))
        {
            printf("MI_IVE_Resize() return ERROR 0x%X\n", s32Ret);
            MI_IVE_Destroy(handle);
            return E_MI_ERR_FAILED;
        }

        // get output from stDst and do something expected.
        ...
    }

    // destroy IVE handle
    MI_IVE_Destroy(handle);
    return MI_SUCCESS;
}

2.41. MI_IVE_Bat¶

Function

Execute the horizontal or vertical alternating time for binary images.

Syntax

MI_S32 MI_IVE_BAT (MI_IVE_HANDLE hHandle, MI_IVE_SrcImage_t *pstSrc, MI_IVE_DstMemInfo_t *pstDstH, MI_IVE_DstMemInfo_t *pstDstV, MVE_IVE_BatCtrl_t *pstCtrl, MI_BOOL bInstant);

Parameter

Parameter Name	Description	Input/Output
hHandle	Regional handle number. Parameter range: [0, MI_IVE_HANDLE_MAX).	Input
pstSrc	Pointer to the source image. It cannot be null.	Input
pstDstH	Pointer to the horizontal output image. It cannot be null.	Output
pstDstV	Pointer to the vertical output image. It cannot be null.	Output
pstCtrl	Pointer to the control parameter. It cannot be null.	Input
bInstant	Reserved.	Input

Parameter Name	Supported Image Type	Address Alignment	Resolution
pstSrc	U8C1 binary image	16 byte	64x64 ~ 1920x1080
pstDstH	U8C1 binary image	16 byte	Same as pstSrc
pstDstV	U8C1 binary image	16 byte	Same as pstSrc

Return Value

Return Value Description

0 Successful.

Non-zero Failed. Please refer to ERROR CODE.
Requirement
- Header: mi_common_datatype.h, mi_ive.h, mi_ive_datatype.h
- Library: libmi_ive.a
Note
- The calculation formula of BAT is as follows:
  
  $O(y)=\left\{ \begin{aligned} 1,h(y)\geqslant u16HorTimes \\ 0,h(y)< u16HorTimes \end{aligned} \right.$
  
  $O(x)=\left\{ \begin{aligned} 1,v(x)\geqslant u16HorTimes \\ 0,v(x)< u16HorTimes \end{aligned} \right.$
  
  Where $h(y)/v(x)$ refer to the horizontal/vertical alternating times, $O(y)/O(x)$ refer to the output image of horizontal/vertical direction.

2.42. MI_IVE_Acc¶

Function

Execute an accumulation task for two gray-scale images.

Syntax

MI_S32 MI_IVE_Acc (MI_IVE_HANDLE hHandle, MI_IVE_SrcImage_t * pstSrc0, MI_IVE_SrcImage_t *pstSrc1, MI_IVE_DstImage_t *pstDst, MVE_IVE_AccCtrl_t * pstAccCtrl, MI_BOOL bInstant);

Parameter

Parameter Name	Description	Input/Output
hHandle	Regional handle number. Parameter range: [0, MI_IVE_HANDLE_MAX).	Input
pstSrc0	Pointer to the source image or the output after processing. It cannot be null.	Input
pstSrc1	Pointer to the source image or the output after processing. It cannot be null.	Input
pstDst	Pointer to the target binary image. It cannot be null.	Output
pstAccCtrl	Pointer to the control parameter. It cannot be null.	Input
bInstant	Reserved.	Input

Parameter Name	Supported Image Type	Address Alignment	Resolution
pstSrc0	U8C1	16 byte	64x64 ~ 1920x1080
pstSrc1	U8C1	16 byte	Same as pstSrc0
pstDst	U8C1	16 byte	Same as pstSrc0

Return Value

Return Value Description

0 Successful.

Non-zero Failed. Please refer to ERROR CODE.
Requirement
- Header: mi_common_datatype.h, mi_ive.h, mi_ive_datatype.h
- Library: libmi_ive.a
Note
- The calculation formulas of different mode are as follows:
  - E_MI_IVE_ACC_MODE_INCREASE :
    
    $I_{out}(x,y)=I_{src0}(x,y)+I_{src1}(x,y)$
  - E_MI_IVE_ACC_MODE_DECREASE :
    
    $I_{out}(x,y)=I_{src0}(x,y)-I_{src1}(x,y)$
  - E_MI_IVE_ACC_MODE_INCREASE_MAP_255TO1 :
    
    $I_{out}(x,y)=I_{src0}(x,y)+(I_{src1}(x,y)\&0x1)$
  Where, $I_{src0}(x,y)$ refer to pstSrc0, $I_{src1}(x,y)$ refer to pstSrc1, $I_{out}(x,y)$ refer to pstDst.

2.43. MI_IVE_Matrix_Transform¶

Function

Execute the operation of matrix multiplication

Syntax

MI_S32 MI_IVE_Matrix_Transform(MI_IVE_HANDLE hHandle, MI_IVE_SrcImage_t *pstSrc1, MI_IVE_SrcImage_t *pstSrc2, MI_IVE_SrcImage_t *pstSrc3, MI_IVE_DstImage_t *pstDst1, MI_IVE_DstImage_t *pstDst2, MI_IVE_DstImage_t *pstDst3, MI_IVE_MatrTranfCtrl_t *pstMatrTranfCtrl, MI_BOOL bInstant);

Parameter

Parameter Name	Description	Input/Output
hHandle	Regional handle number. Parameter range: [0, MI_IVE_HANDLE_MAX).	Input
pstSrc1	Pointer to the source image. It cannot be null.	Input
pstSrc2	Pointer to the source image. It cannot be null.	Input
pstSrc3	Pointer to the source image. It cannot be null.	Input
pstDst1	Pointer to the output image. It cannot be null.	Output
pstDst2	Pointer to the output image. It cannot be null.	Output
pstDst3	Pointer to the output image. It cannot be null.	Output
pstMatrTranfCtrl	Pointer to the control parameter. It cannot be null.	Input
bInstant	Reserved.	Input

Parameter Name	Supported Image Type	Address Alignment	Resolution
pstSrc1	S32C1	16 byte	64x64 ~ 1920x1080
pstSrc2	S32C1	16 byte	Same as pstSrc1
pstSrc3	S32C1	16 byte	Same as pstSrc1
pstDst1	S32C1	16 byte	Same as pstSrc1
pstDst2	S32C1	16 byte	Same as pstSrc1
pstDst3	S32C1	16 byte	Same as pstSrc1

Return Value

Return Value Description

0 Successful.

Non-zero Failed. Please refer to ERROR CODE.
Requirement
- Header: mi_common_datatype.h, mi_ive.h, mi_ive_datatype.h
- Library: libmi_ive.a
Note
- The matrix_transform calculation formula is as illustrated in the following figure.
  
  Fig 1-13: Illustration of Matrix_transform Calculation Formula
  
  Each input and output element consists of 16 integer, 15 decimal and 1 sign.
  
  Each input element of matrix consists of 16 integer, 15 decimal and 1 sign.
  - For Pudding series chips, each input element of matrix consists of 1 integer, 14 decimal and 1 sign.
  - The input type should be S32C1, for the common standard of different chip format.

2.44. MI_IVE_Image_Dot¶

Function

Execute the operation of dot product.

Syntax

MI_S32 MI_IVE_Image_Dot (MI_IVE_HANDLE hHandle, MI_IVE_SrcImage_t *pstSrc1, MI_IVE_SrcImage_t *pstSrc2, MI_IVE_DstImage_t *pstDst, MI_BOOL bInstant);

Parameter

Parameter Name	Description	Input/Output
hHandle	Regional handle number. Parameter range: [0, MI_IVE_HANDLE_MAX).	Input
pstSrc1	Pointer to the source image. It cannot be null.	Input
pstSrc2	Pointer to the source image. It cannot be null.	Input
pstDst	Pointer to the output image. It cannot be null.	Output
bInstant	Reserved.	Input

Parameter Name	Supported Image Type	Address Alignment	Resolution
pstSrc1	S32C1	16 byte	64x64 ~ 1920x1080
pstSrc2	S32C1	16 byte	Same as pstSrc1
pstDst	S32C1	16 byte	Same as pstSrc1

Return Value

Return Value Description

0 Successful.

Non-zero Failed. Please refer to ERROR CODE.
Requirement
- Header: mi_common_datatype.h, mi_ive.h, mi_ive_datatype.h
- Library: libmi_ive.a
Note
- The calculation formula of Image Dot is as follows:
  
  $I_{out}(x,y) = I_{src1}(x,y) \times I_{src2}(x,y)$
  
  Where $I_{src1}(x,y)$ refer to pstSrc1, $I_{src2}(x,y)$ refer to pstSrc2, $I_{out}(x,y)$ refer to pstDst.
  
  Each input and output element consists of 16 integer, 15 decimal and 1 sign.

2.45. MI_IVE_Shift_Detector¶

Function

Execute the operation of object tracking.

Syntax

MI_S32 MI_IVE_Shift_Detector(MI_IVE_HANDLE hHandle, MI_IVE_SrcImage_t *pstSrc1, MI_IVE_SrcImage_t *pstSrc2, MI_IVE_DstImage_t MI_IVE_DstImage_t *pstDstX, MI_IVE_DstImage_t *pstDstY, MI_IVE_SHIFT_DETECT_CTRL_t *pstCtrl, MI_BOOL bInstant);

Parameter

Parameter Name	Description	Input/Output
hHandle	Regional handle number. Parameter range: [0, MI_IVE_HANDLE_MAX).	Input
pstSrc0	Pointer to the first source image. It cannot be null.	Input
pstSrc1	Pointer to the second source image. It cannot be null.	Input
pstDstX	Pointer to the shift x image. It cannot be null.	Output
pstDstY	Pointer to the shift y image. It cannot be null.	Output
pstCtrl	Pointer to the control parameter It cannot be null.	Input
bInstant	Reserved.	Input

Parameter Name	Supported Image Type	Address Alignment	Resolution
pstSrc1	U8C1	16 byte	64x64 ~ 1920x1080
pstSrc2	U8C1	16 byte	Same as pstSrc1
pstDstX	U8C1	16 byte	Same as pstSrc1
pstDstY	U8C1	16 byte	Same as pstSrc1

Return Value

Return Value Description

0 Successful.

Non-zero Failed. Please refer to ERROR CODE.
Requirement
- Header: mi_common_datatype.h, mi_ive.h, mi_ive_datatype.h
- Library: libmi_ive.a
Note
- Enter two images (Tk and Tk+1) of the same scene but in different time series, and input the positions of Tk target area by controlling data pointer, and the shift detector would output the tracing result of the target area in Tk+1.
- If enMode = E_MI_IVE_SHIFT_DETECT_MODE_MULTI, the shift detector would divide images into MxN blocks (block size = u16Width x u16Height) and track all objects in each block.
- If enMode = E_MI_IVE_SHIFT_DETECT_MODE_SINGLE, the shift detector would only track single object on images.

2.46. MI_IVE_AlphaBlending¶

Function

Execute an independent weighted addition calculation task against two grayscale images.

Syntax

MI_S32 MI_IVE_AlphaBlending (MI_IVE_HANDLE hHandle, MI_IVE_SrcImage_t *pstSrc1, MI_IVE_SrcImage_t *pstSrc2, MI_IVE_SrcImage_t *pstAlpha, MI_IVE_DstImage_t *pstDst, MI_BOOL bInstant);

Parameter

Parameter Name	Description	Input/Output
hHandle	Regional handle number. Parameter range: [0, MI_IVE_HANDLE_MAX).	Input
pstSrc1	Pointer to the source image. It cannot be null.	Input
pstSrc2	Pointer to the source image. It cannot be null.	Input
pstAlpha	Pointer to the alpha map. It cannot be null.	Input
pstDst	Pointer to the output image. It cannot be null.	Output
bInstant	Reserved.	Input

Parameter Name	Supported Image Type	Address Alignment	Resolution
pstSrc1	U8C1	16 byte	64x64 ~ 1920x1080
pstSrc2	U8C1	16 byte	Same as pstSrc1
pstAlpha	U8C1	16 byte	Same as pstSrc1
pstDst	U8C1	16 byte	Same as pstSrc1

Return Value

Return Value Description

0 Successful.

Non-zero Failed. Please refer to ERROR CODE.
Requirement
- Header: mi_common_datatype.h, mi_ive.h, mi_ive_datatype.h
- Library: libmi_ive.a
Note
- The calculation formula of AlphaBlending is as follows:
  
  $I_{\text{Out}} = I_{Src1}\left( x,y \right) \times I_{\text{Alpha}}(x,y) + I_{Src2}(x,y) \times (255 - I_{\text{Alpha}}(x,y))$
  
  Where $I_{src1}(x,y)$ refer to pstSrc1, $I_{src2}(x,y)$ refer to pstSrc2, $I_{Alpha}(x,y)$ refer to pstAlpha, $I_{out}(x,y)$ refer to pstDst.

2.47. MI_IVE_BGBlur¶

Function

Execute the task of blur, replacement, mosaic or blur with mosaic for image background.

Syntax

MI_S32 MI_IVE_BGBlur(MI_IVE_HANDLE hHandle, MI_IVE_SrcImage_t *pstSrcYMask, MI_IVE_SrcImage_t *pstSrcUvMask, MI_IVE_SrcImage_t *pstSrcOri, MI_IVE_SrcImage_t *pstSrcRepBg, MI_IVE_DstImage_t *pstDst, MI_IVE_BgBlurCtrl_t *pstCtrl);

Parameter

Parameter Name	Description	Input/Output
hHandle	Regional handle number. Parameter range: [0, MI_IVE_HANDLE_MAX).	Input
pstSrcYMask	Pointer to the source image which labeled the foreground/background of the Y component of pstSrcOri. It cannot be null.	Input
pstSrcUvMask	Pointer to the source image which labeled the foreground/background of the UV component of pstSrcOri. It cannot be null.	Input
pstSrcOri	Pointer to the source image. It cannot be null.	Input
pstSrcRepBg	Pointer to the source image. Please refer to the tabulation below. It cannot be null.	Input
pstDst	Pointer to the output image. It cannot be null.	Output
pstCtrl	Pointer to the control parameter. It cannot be null.	Input

Parameter declaration of Souffle series

Parameter Name	Supported Image Type	Address Alignment	Resolution
pstSrcYMask	U8C1	16 byte	640x360 or 1280x720
pstSrcUvMask	U8C1	16 byte	320x180 or 640x360, should be corresponding to pstSrcYMask
pstSrcOri	YUV420SP, YUV422_YUYV	16 byte	640x360~7680x4320, the width can be 2-pixel-align.
pstSrcRepBg	YUV420SP for background blur. YUV420SP, YUV422_YUYV for background replacement.	16 byte	Same as pstSrcYMask for background blur. 64x64~7680x4320 for background replacement, the width can be 2-pixel-align.
pstDst	YUV420SP, YUV422_YUYV	16 byte	Same as pstSrcOri

Parameter declaration of iFord series

Parameter Name	Supported Image Type	Address Alignment	Resolution
pstSrcYMask	U8C1	16 byte	The width is 640 or 480 The height is ALIGN_DOWN(pstSrcYMask->u16Width * pstSrcOri->u16Height / pstSrcOri->u16Width, 1)
pstSrcUvMask	U8C1	16 byte	The width/height is half of the width/height of pstSrcYMask.
pstSrcOri	YUV420SP	16 byte	640x360~3840x2160, the width can be 2-pixel-align.
pstSrcRepBg	YUV420SP	16 byte	Same as pstSrcYMask.
pstDst	YUV420SP	16 byte	Same as pstSrcOri.

For iFord series, the width of pstSrcYMask is 640 when pstSrcOri->u16Width >= pstSrcOri->u16Height, 480 when pstSrcOri->u16Width < pstSrcOri->u16Height.
ALIGN_DOWN(n, 1) indicates that n is aligned down to 2, e.g. ALIGN_DOWN(513, 1) = 512

Return Value

Return Value Description

0 Successful.

Non-zero Failed. Please refer to ERROR CODE.
Requirement
- Header: mi_common_datatype.h, mi_ive.h, mi_ive_datatype.h
- Library: libmi_ive.a

Note

The BGBlur mode, please refer to MI_IVE_BgBlurMode_e .

BGBlur mode	The limitation of the source image size
Background blur/Background Mosaic/Background blur and mosaic	pstSrcYMask = pstSrcRepBg <= pstSrcOri
Background replacement	pstSrcYMask <= pstSrcOri; pstSrcOri / 16 <= pstSrcRepBg <= pstSrcOri NOTE: the resolution of pstSrcRepBg should be the same as pstSrcOri when the width of pstSrcRepBg is over 1280.

Example

MI_S32 Sample_BGBlur()
{
    MI_S32 s32Ret;
    MI_IVE_HANDLE handle = 0;
    MI_IVE_SrcImage_t stSrcOri, stSrcYMask, stSrcUvMask, stSrcRepBg;
    MI_IVE_DstImage_t stDst;
    MI_IVE_BgBlurCtrl_t ctrl =
    {
        .eBgBlurMode    = E_MI_IVE_BGBLUR_MODE_BLUR, // or other mode.
        .u8MaskThr      = 127, // [0, 255], depends on user.
        .u8BlurLv       = 0, // [0, 255], depends on user and only valid in Souffle series.
        .u8ScalingStage = 0, // [0, 255], used in E_MI_IVE_BGBLUR_MODE_BLUR.
        .eBgBlurMaskOp  = E_MI_IVE_BGBLUR_MASK_OP_DILATE, // depends on user, refer to MI_IVE_BgBlurMaskOp_e.
        .u8SaturationLv = 64, // [0, 128], depends on user and only iFord support.
        .u8MosaicSize   = 6, // [2, 4, 6, 8, 10], depends on user and only iFord support.
    };

    // Init IVE
    if (MI_SUCCESS != (s32Ret = MI_IVE_Create(handle)))
    {
        printf("Could not create IVE handle\n");
        break;
    }

    // ...set input image info to stSrcOri, stSrcYMask, stSrcUvMask, stSrcRepBg.
    ...
    // ...set output image info to stDst.
    ...

    // BGBlur task.
    if (MI_SUCCESS != (s32Ret = MI_IVE_BGBlur(handle, &stSrcYMask, &stSrcUvMask, &stSrcOri, &stSrcRepBg, &stDst, &ctrl)))
    {
        printf("MI_IVE_BGBlur() return ERROR 0x%X\n", s32Ret);
        return E_MI_ERR_FAILED;
    }

    // get output from stDst and do something expected.
    ...

    // Deinit IVE
    MI_IVE_Destroy(handle);

    return s32Ret;
}

Fig 1-14: Illustration of Background Blur.

Fig 1-15: Illustration of Background Replacement.

3. DATA TYPE¶

3.1. Data Type List¶

The table below lists the data structure definitions of the related IVE data types:

Data type	Definition
MI_IVE_HANDLE_MAX	Defines the maximum number of IVE handle.
MI_IVE_HIST_NUM	Defines histogram statistics bin number
MI_IVE_MAP_NUM	Defines number of mapping lookup table entries
MI_IVE_MAX_REGION_NUM	Defines maximum number of connected regions
MI_IVE_ST_MAX_CORNER_NUM	Defines maximum number of Shi-Tomasi-like corners
MI_IVE_MASK_SIZE_5X5	Defines Mask size
MI_IVE_CANNY_STACK_RESERVED_SIZE	Defines Canny Stack Reserved Size
MI_IVE_ImageType_e	Defines image types supported by generalized two-dimensional images
MI_IVE_Image_t	Defines generalized two-dimensional image information
MI_IVE_SrcImage_t	Defines source image
MI_IVE_DstImage_t	Defines output image
MI_IVE_Data_t	Defines two-dimensional image information in byte unit
MI_IVE_SrcData_t	Defines two-dimensional source data information in byte unit
MI_IVE_DstData_t	Defines two-dimensional output data information in byte unit
MI_IVE_MemInfo_t	Defines one-dimensional data memory information
MI_IVE_SrcMemInfo_t	Defines one-dimensional source data
MI_IVE_DstMemInfo_t	Defines one-dimensional output data
MI_IVE_Length8bit_u	Defines 8-bit data union
MI_IVE_PointU16_t	Defines U16 point information structure
MI_IVE_PointS25Q7_t	Defines S25Q7 fixed point information structure
MI_IVE_Rect_t	Defines U16 rectangle information structure
MI_IVE_FilterCtrl_t	Defines template filter control information
MI_IVE_CscMode_e	Defines color space conversion mode
MI_IVE_CscCtrl_t	Defines color space conversion control information
MI_IVE_FilterAndCscCtrl_t	Defines template filter plus color space conversion composite function control information
MI_IVE_SobelOutCtrl_e	Defines Sobel output control information
MI_IVE_SobelCtrl_t	Defines Sobel edge extraction control information
MI_IVE_MagAndAngOutCtrl_e	Defines Canny edge magnitude and angle calculation output format
MI_IVE_MagAndAngCtrl_t	Defines Canny edge magnitude and angle calculation control information
MI_IVE_DilateCtrl_t	Defines Dilate control information
MI_IVE_ErodeCtrl_t	Defines Erode control information
MI_IVE_ThreshMode_e	Defines image thresholding output format
MI_IVE_ThreshCtrl_t	Defines image thresholding control information
MI_IVE_SubMode_e	Defines image subtraction output format
MI_IVE_SubCtrl_t	Defines image subtraction control parameter
MI_IVE_IntegOutCtrl_e	Defines integral map output control parameter
MI_IVE_IntegCtrl_t	Defines integral map calculation control parameter
MI_IVE_ThreshS16Mode_e	Defines 16-bit signed image thresholding mode
MI_IVE_ThreshS16Ctrl_t	Defines 16-bit signed image thresholding control parameter
MI_IVE_ThreshU16Mode_e	Defines 16-bit unsigned image thresholding mode
MI_IVE_ThreshU16Ctrl_t	Defines 16-bit unsigned image thresholding control parameter
MI_IVE_16BitTo8BitMode_e	Defines 16-bit image to 8-bit image conversion mode
MI_IVE_16bitTo8BitCtrl_t	Defines 16-bit image to 8-bit image conversion control parameter
MI_IVE_OrdStatFilterMode_e	Defines sequential statistic filtering mode
MI_IVE_OrdStatFilter_t	Defines sequential statistic filtering control parameter
MI_IVE_MapLutMem_t	Defines Map operator lookup table memory information
MI_IVE_EqualizeHistCtrlMem_t	Defines histogram equalization auxiliary memory
MI_IVE_EqualizeHistCtrl_t	Defines histogram equalization control parameter
MI_IVE_AddMode_e	Define the add calculation mode
MI_IVE_AddCtrl_t	Defines image weighted addition control parameter
MI_IVE_NccDstMem_t	Defines NCC output memory information
MI_IVE_Region_t	Defines connected region information
MI_IVE_CcBlob_t	Defines connected region label output information
MI_IVE_CclMode_e	Defines connected region mode
MI_IVE_CclCtrl_t	Defines connected region label control parameter
MI_IVE_GmmCtrl_t	Defines GMM background modelling control parameter
MI_IVE_CannyStackSize_t	Defines strong edge point stack size structure in Canny edge first-half calculation
MI_IVE_CannyHysEdgeCtrl_t	Defines control parameter in Canny edge first-half calculation
MI_IVE_LbpCmpMode_e	Defines LBP calculation comparison mode.
MI_IVE_LbpChalMode_e	Define LBP input channel mode
MI_IVE_LbpCtrrl_t	Defines LBP texture calculation control parameter
MI_IVE_NormGradOutCtrl_e	Defines normalized gradient information calculation task output control enumeration type
MI_IVE_NormGradCtrl_t	Defines normalized gradient information calculation control parameter
MI_IVE_MvS9Q7_t	Defines displacement structure
MI_IVE_LkOpticalFlowCtrl_t	Defines LK optical flow calculation control parameter
MI_IVE_SadMode_e	Defines SAD calculation mode
MI_IVE_SadOutCtrl_e	Defines SAD output control mode
MI_IVE_SadCtrl_t	Defines SAD control parameter
MI_IVE_BernsenCtrl_t	Defines Bernsen thresh control parameters.
MI_IVE_BernsenMode_e	Defines the Bernsen thresh mode.
MI_IVE_LineFilterHorCtrl_t	Defines control parameters for filtering the horizontal density of binary images.
MI_IVE_LineFilterVerCtrl_t	Defines control parameters for filtering the vertical density of binary images.
MI_IVE_NoiseRemoveHor_t	Defines the horizontal noise removal control parameter for the binary image.
MI_IVE_NoiseRemoveVer_t	Defines the vertical noise removal control parameter for the binary image.
MI_IVE_AdpThreshCtrl_t	Defines adaptive thresh control parameters.
MI_IVE_ResizeMode_e	Defines the resize input image mode
MI_IVE_ResizeMethod_e	Defines the resize method.
MI_IVE_ResizeCtrl_t	Defines the resize control parameter
MI_IVE_BatCtrl_t	Defines the bat control parameter
MI_IVE_BatMode_e	Defines the bat output control mode
MI_IVE_AccCtrl_t	Defines the acc control parameter
MI_IVE_AccMode_e	Defines the acc input control mode
MI_IVE_MatrTranfMode_e	Defines the input channel mode of matrix transform
MI_IVE_MatrTranfCtrlMode_e	Defines the input control mode of matrix transform.
MI_IVE_MatrTranfCtrl_t	Defines the control parameters of matrix transform.
MI_IVE_SHIFT_DETECT_MODE_e	Defines the shift detector input control mode.
MI_IVE_SHIFT_DETECT_CTRL_t	Defines control parameter of the shift detector.
MI_IVE_BgBlurMode_e	Defines the BGBlur mode.
MI_IVE_BgBlurMaskOp_e	Define the Mask Preprocess mode of BGBlur.
MI_IVE_BgBlurCtrl_t	Defines the control parameters of BGBlur.

3.2. Fixed Point Data Type¶

Description

Defines fixed point data type.

Definition

typedef unsigned char   MI_U0Q8;
typedef unsigned char   MI_U1Q7;
typedef unsigned char   MI_U5Q3;
typedef unsigned short  MI_U0Q16;
typedef unsigned short  MI_U4Q12;
typedef unsigned short  MI_U6Q10;
typedef unsigned short  MI_U8Q8;
typedef unsigned short  MI_U14Q2;
typedef unsigned short  MI_U12Q4;
typedef short           MI_S14Q2;
typedef short           MI_S9Q7;
typedef unsigned int    MI_U22Q10;
typedef unsigned int    MI_U25Q7;
typedef int             MI_S25Q7;
typedef unsigned short  MI_U8Q4F4; /*8bits unsigned integer, 4bits decimal fraction, 4bits flag bits*/

Member

Member Code	Description
MI_U0Q8	0 bit for the integer part, and 8 bits for the decimal part. Represented in document as UQ0.8.
MI_U1Q7	High-order 1-bit unsigned data for the integer part, and low-order 7 bits for the demical part. Represented in document as UQ1.7.
MI_U5Q3	High-order 5-bit unsigned data for the integer part, and low-order 3 bits for the demical part. Represented in document as UQ5.3.
MI_U0Q16	0 bit for the integer part, and 16 bits for the demical part. Represented in document as UQ0.16.
MI_U4Q12	High-order 4-bit unsigned data for the integer part, and low-order 12 bits for the demical part. Represented in document as UQ4.12.
MI_U6Q10	High-order 6-bit unsigned data for the integer part, and low-order 10 bits for the demical part. Represented in document as UQ6.10.
MI_U8Q8	High-order 8-bit unsigned data for the integer part, and low-order 8 bits for the demical part. Represented in document as UQ8.8.
MI_U14Q2	High-order 14-bit unsigned data for the integer part, and low-order 2 bits for the demical part. Represented in document as UQ14.2.
MI_U12Q4	High-order 12-bit unsigned data for the integer part, and low-order 4 bits for the demical part. Represented in document as UQ12.4.
MI_S14Q2	High-order 14-bit signed data for the integer part, and low-order 2 bits for the demical part. Represented in document as SQ14.2.
MI_S9Q7	High-order 9-bit signed data for the integer part, and low-order 7 bits for the demical part. Represented in document as SQ9.7.
MI_U22Q10	High-order 22-bit unsigned data for the integer part, and low-order 10 bits for the demical part. Represented in document as UQ22.10.
MI_U25Q7	High-order 25-bit unsigned data for the integer part, and low-order 7 bits for the demical part. Represented in document as UQ25.7.
MI_S25Q7	High-order 25-bit signed data for the integer part, and low-order 7 bits for the demical part. Represented in document as SQ25.7.
MI_U8Q4F4	High-order 8-bit unsigned data for the integer part, middle 4 bits for the demical part, and low-order 4 bits for flagging. Represented in document as UQF8.4.4.

Note
- MI_UxQyFz/MI_SxQy:
  - The number x after the letter U means x bits of unsigned data are used for the integer part.
  - The number x after the letter S means x bits of signed data are used for the integer part.
  - The number y after the letter Q means y bits of data are used for the demical part.
  - The number z after the letter F means z bits are used for flagging.
  - Highest bit on the leftmost side and lowest bit on the rightmost side.

3.3. MI_IVE_HANDLE_MAX¶

Description

Defines the maximum number of IVE handle.
Definition
```
#define MI_IVE_HANDLE_MAX 128
```

3.4. MI_IVE_HIST_NUM¶

Description

Defines histogram statistics bin number.
Definition
```
#define MI_IVE_HIST_NUM 256
```

3.5. MI_IVE_MAP_NUM¶

Description

Defines number of mapping lookup table entries.
Definition
```
#define MI_IVE_MAP_NUM 256
```

3.6. MI_IVE_MAX_REGION_NUM¶

Description

Defines maximum number of connected regions.
Definition
```
#define MI_IVE_MAX_REGION_NUM 255
```

3.7. MI_IVE_ST_MAX_CORNER_NUM¶

Description

Defines maximum number of Shi-Tomasi-like corners.
Definition
```
#define MI_IVE_ST_MAX_CORNER_NUM 200
```

3.8. MI_IVE_MASK_SIZE_5X5¶

Description

Defines mask size.
Definition
```
#define MI_IVE_MASK_SIZE_5X5 25
```

3.9. MI_IVE_CANNY_STACK_RESERVED_SIZE¶

Description

Defines Canny stack reserved size.

Definition

#define MI_IVE_CANNY_STACK_RESERVED_SIZE 12

3.10. MI_IVE_ImageType_e¶

Description

Defines image types supported by generalized two-dimensional images.

Definition

typedef enum
{
    E_MI_IVE_IMAGE_TYPE_U8C1          =0x0,
    E_MI_IVE_IMAGE_TYPE_S8C1          =0x1,
    E_MI_IVE_IMAGE_TYPE_YUV420SP      =0x2, /*YUV420 SemiPlanar*/
    E_MI_IVE_IMAGE_TYPE_YUV422SP      =0x3, /*YUV422 SemiPlanar*/
    E_MI_IVE_IMAGE_TYPE_YUV420P       =0x4, /*YUV420 Planar*/
    E_MI_IVE_IMAGE_TYPE_YUV422P       =0x5, /*YUV422 planar*/
    E_MI_IVE_IMAGE_TYPE_S8C2_PACKAGE  =0x6,
    E_MI_IVE_IMAGE_TYPE_S8C2_PLANAR   =0x7,
    E_MI_IVE_IMAGE_TYPE_S16C1         =0x8,
    E_MI_IVE_IMAGE_TYPE_U16C1         =0x9,
    E_MI_IVE_IMAGE_TYPE_U8C3_PACKAGE  =0xa,
    E_MI_IVE_IMAGE_TYPE_U8C3_PLANAR   =0xb,
    E_MI_IVE_IMAGE_TYPE_S32C1         =0xc,
    E_MI_IVE_IMAGE_TYPE_U32C1         =0xd,
    E_MI_IVE_IMAGE_TYPE_S64C1         =0xe,
    E_MI_IVE_IMAGE_TYPE_U64C1         =0xf,
    E_MI_IVE_IMAGE_TYPE_YUV420SP_NV21 =0x10,
    E_MI_IVE_IMAGE_TYPE_YUV422_YUYV   =0x11,
    E_MI_IVE_IMAGE_TYPE_BUTT
}MI_IVE_ImageType_e;

Member

Member Code	Description
E_MI_IVE_IMAGE_TYPE_U8C1	Single-channel image, in which each pixel is represented by one 8-bit unsigned data entry.
E_MI_IVE_IMAGE_TYPE_S8C1	Single-channel image, in which each pixel is represented by one 8-bit signed data entry.
E_MI_IVE_IMAGE_TYPE_YUV420SP	Image in YUV420 semiplanar format.
E_MI_IVE_IMAGE_TYPE_YUV422SP	Image in YUV422 semiplanar format.
E_MI_IVE_IMAGE_TYPE_YUV420P	Image in YUV420 planar format.
E_MI_IVE_IMAGE_TYPE_YUV422P	Image in YUV422 planar format.
E_MI_IVE_IMAGE_TYPE_S8C2_PACKAGE	Dual-channel image in package format, in which each pixel is represented by two 8-bit signed data entries.
E_MI_IVE_IMAGE_TYPE_S8C2_PLANAR	Dual-channel image in planar format, in which each pixel is represented by two 8-bit signed data entries.
E_MI_IVE_IMAGE_TYPE_S16C1	Single-channel image, in which each pixel is represented by one 16-bit signed data entry.
E_MI_IVE_IMAGE_TYPE_U16C1	Single-channel image, in which each pixel is represented by one 16-bit unsigned data entry.
E_MI_IVE_IMAGE_TYPE_U8C3_PACKAGE	3-channel image in package format, in which each pixel is represented by three 8-bit unsigned data entries.
E_MI_IVE_IMAGE_TYPE_U8C3_PLANAR	3-channel image in planar format, in which each pixel is represented by three 8-bit unsigned data entries.
E_MI_IVE_IMAGE_TYPE_S32C1	Single-channel image, in which each pixel is represented by one 32-bit signed data entry.
E_MI_IVE_IMAGE_TYPE_U32C1	Single-channel image, in which each pixel is represented by one 32-bit unsigned data entry.
E_MI_IVE_IMAGE_TYPE_S64C1	Single-channel image, in which each pixel is represented by one 64-bit signed data entry.
E_MI_IVE_IMAGE_TYPE_U64C1	Single-channel image, in which each pixel is represented by one 64-bit unsigned data entry.
E_MI_IVE_IMAGE_TYPE_YUV420SP_NV21	Image in YUV420 semiplanar format, the placement of UV component is opposite to E_MI_IVE_IMAGE_TYPE_YUV420SP.
E_MI_IVE_IMAGE_TYPE_YUV422_YUYV	Image in YUV422 package format which placed data in order of YUYV.

Related Data Type and Interface

MI_IVE_Image_t

MI_IVE_SrcImage_t

MI_IVE_DstImage_t

3.11. MI_IVE_Image_t¶

Description

Defines generalized two-dimensional image information.

Definition

typedef struct MI_IVE_Image_s
{
    MI_IVE_ImageType_e eType;
    MI_PHY aphyPhyAddr[3];
    MI_U8 *apu8VirAddr[3];
    MI_U16 au16Stride[3];
    MI_U16 u16Width;
    MI_U16 u16Height;
    MI_U16 u16Reserved; /*Can be used such as elemSize*/
}MI_IVE_Image_t;

Member

Member Code	Description
enType	Generalized image type.
aphyPhyAddr[3]	Generalized image physical address array.
apu8VirAddr[3]	Generalized image virtual address array.
au16Stride[3]	Generalized image stride.
u16Width	Generalized image width.
u16Height	Generalized image height.
u16Reserved	Mark the memory as being used by software or hardware to determine the operation of automatic Cache Invalidate and Cache Flush.

Note
- Different operators have different requirements as to the alignment of image input/output address.
- u16Width, u16Height and u16Stride are all in pixel unit.
Related Data Type and Interface

MI_IVE_ImageType_e

MI_IVE_SrcImage_t

MI_IVE_DstImage_t

3.12. MI_IVE_SrcImage_t¶

Description

Defines source image.

Definition

typedef MI_IVE_Image_t MI_IVE_SrcImage_t;

Related Data Type and Interface

MI_IVE_Image_t

MI_IVE_DstImage_t

3.13. MI_IVE_DstImage_t¶

Description

Defines output image.

Definition

typedef MI_IVE_Image_t MI_IVE_DstImage_t;

Related Data Type and Interface

MI_IVE_Image_t

MI_IVE_SrcImage_t

3.14. MI_IVE_Data_t¶

Description

Defines two-dimensional data information in byte unit.

Definition

typedef struct MI_IVE_Data_s
{
    MI_PHY phyPhyAddr; /*Physical address of the data*/
    MI_U8 *pu8VirAddr;
    MI_U16 u16Stride; /*Data stride by byte*/
    MI_U16 u16Height; /*Data height by byte*/
    MI_U16 u16Width; /*Data width by byte*/
    MI_U16 u16Reserved;
}MI_IVE_Data_t;

Member

Member Code	Description
phyPhyAddr	Image physical address.
pu8VirAddr	Image virtual address.
u16Stride	Image stride.
u16Height	Image width.
u16Width	Image height.
u16Reserved	Reserved.

Note
- Represents two-dimensional data in byte unit; can be converted with MI_IVE_Image_t image.

3.15. MI_IVE_SrcData_t¶

Description

Defines two-dimensional source data information in byte unit

Definition

typedef MI_IVE_Data_t MI_IVE_SrcData_t;

Related Data Type and Interface

MI_IVE_Image_t

MI_IVE_DstData_t

3.16. MI_IVE_DstData_t¶

Description

Defines two-dimensional output data information in byte unit.

Definition

typedef MI_IVE_Data_t MI_IVE_DstData_t;

Related Data Type and Interface

MI_IVE_Image_t

MI_IVE_SrcImage_t

3.17. MI_IVE_MemInfo_t¶

Description

Defines one-dimensional data memory information.

Definition

typedef struct MI_IVE_MemInfo_s
{
    MI_PHY phyPhyAddr;
    MI_U8 *pu8VirAddr;
    MI_U32 u32Size;
}MI_IVE_MemInfo_t;

Member

Member Code Description

phyPhyAddr One-dimensional data physical address.

pu8VirAddr One-dimensional data virtual address.

u32Size One-dimensional data byte size.
Related Data Type and Interface

MI_IVE_SrcMemInfo_t

MI_IVE_DstMemInfo_t

3.18. MI_IVE_SrcMemInfo_t¶

Description

Defines one-dimensional source data.

Definition

typedef MI_IVE_MemInfo_t MI_IVE_SrcMemInfo_t;

Related Data Type and Interface

MI_IVE_MemInfo_t

MI_IVE_DstMemInfo_t

3.19. MI_IVE_DstMemInfo_t¶

Description

Defines one-dimensional output data.

Definition

typedef MI_IVE_MemInfo_t MI_IVE_DstMemInfo_t;

Related Data Type and Interface

MI_IVE_MemInfo_t

MI_IVE_SrcMemInfo_t

3.20. MI_IVE_Length8bit_u¶

Description

Defines 8-bit data union.

Definition

typedef union
{
    MI_S8 s8Val;
    MI_U8 u8Val;
} MI_IVE_Length8bit_u;

Member

Member Code Description

s8Val Signed 8-bit value.

u8Val Unsigned 8-bit value.

3.21. MI_IVE_PointU16_t¶

Description

Defines U16 point information structure.

Definition

typedef struct MI_IVE_PointU16_s
{
    MI_U16 u16X;
    MI_U16 u16Y;
}MI_IVE_PointU16_t;

Member

Member Code Description

u16X x coordinate of point.

u16Y y coordinate of point.

3.22. MI_IVE_PointS25Q7_t¶

Description

Defines S25Q7 fixed-point information structure.

Definition

typedef struct MI_IVE_PointS25Q7_s
{
    MI_S25Q7 s25q7X; /*X coordinate*/
    MI_S25Q7 s25q7Y; /*Y coordinate*/
}MI_IVE_PointS25Q7_t;

Member

Member Code Description

s25q7X x coordinate of point, in SQ25.7 representation.

s25q7Y y coordinate of point, in SQ25.7 representation.

3.23. MI_IVE_Rect_t¶

Description

Defines U16 rectangle information structure.

Definition

typedef struct MI_IVE_Rect_s
{
    MI_U16 u16X;
    MI_U16 u16Y;
    MI_U16 u16Width;
    MI_U16 u16Height;
}MI_IVE_Rect_t;

Member

Member Code	Description
u16X	X coordinate of the rectangle closest to the coordinate origin.
u16Y	Y coordinate of the rectangle closest to the coordinate origin.
u16Width	Width of rectangle.
u16Height	Height of rectangle.

3.24. MI_IVE_FilterCtrl_t¶

Description

Defines template filter control information.

Defines

typedef struct MI_IVE_FilterCtrl_s
{
    MI_S8 as8Mask[MI_IVE_MASK_SIZE_5X5]; /*Template parameter filter
    coefficient*/
    MI_U8 u8Norm; /*Normalization parameter, by right shift*/
}MI_IVE_FilterCtrl_t;

Member

Member Code	Description
as8Mask[MI_IVE_MASK_SIZE_5X5]	5x5 template coefficient. Setting the peripheral coefficient to 0 can realize a 3x3 template filter, configured in row-major order.
u8Norm	Normalization parameter. Parameter range: [0, 13].

Note
- By configuring different template coefficients, you can obtain different filter effects.

3.25. MI_IVE_CscMode_e¶

Description

Defines color space conversion mode.

Definition

typedef enum
{
    E_MI_IVE_CSC_MODE_PIC_BT601_YUV2RGB      = 0x0,
    E_MI_IVE_CSC_MODE_PIC_BT601_YUV2BGR      = 0x1,
    E_MI_IVE_CSC_MODE_PIC_BT601_RGB2YUV      = 0x2,
    E_MI_IVE_CSC_MODE_PIC_BT601_BGR2YUV      = 0x3,
    E_MI_IVE_CSC_MODE_PIC_BT601_NV21_YUV2RGB = 0x4,
    E_MI_IVE_CSC_MODE_PIC_BT601_NV21_YUV2BGR = 0x5,
    E_MI_IVE_CSC_MODE_PIC_BT601_RGB2YUV_NV21 = 0x6,
    E_MI_IVE_CSC_MODE_PIC_BT601_BGR2YUV_NV21 = 0x7,
    E_MI_IVE_CSC_MODE_MAX
}MI_IVE_CscMode_e;

Member

Member Code	Description
E_MI_IVE_CSC_MODE_PIC_BT601_YUV2RGB	YUV(NV12)2RGB image conversion.
E_MI_IVE_CSC_MODE_PIC_BT601_YUV2BGR	YUV(NV12)2BGR image conversion.
E_MI_IVE_CSC_MODE_PIC_BT601_RGB2YUV	RGB2YUV(NV12) image conversion.
E_MI_IVE_CSC_MODE_PIC_BT601_BGR2YUV	BGR2YUV(NV12) image conversion.
E_MI_IVE_CSC_MODE_PIC_BT601_NV21_YUV2RGB	YUV(NV21)2RGB image conversion.
E_MI_IVE_CSC_MODE_PIC_BT601_NV21_YUV2BGR	YUV(NV21)2BGR image conversion.
E_MI_IVE_CSC_MODE_PIC_BT601_RGB2YUV_NV21	RGB2YUV(NV21) image conversion.
E_MI_IVE_CSC_MODE_PIC_BT601_BGR2YUV_NV21	BGR2YUV(NV21) image conversion.

Note
- Input and output data must satisfy 0≤Y, U, V≤255.
- Input and output data must satisfy 0≤R, G, B≤255.
Related Data Type and Interface

MI_IVE_CscCtrl_t

MI_IVE_FilterAndCscCtrl_t

3.26. MI_IVE_CscCtrl_t¶

Description

Defines color space conversion control information.

Definition

typedef struct MI_IVE_CscCtrl_s
{
    MI_IVE_CscMode_e enMode; /*Working mode*/
}MI_IVE_CscCtrl_t;

Member

Member Code Description

enMode Work mode.
Related Data Type and Interface

MI_IVE_CscMode_e

3.27. MI_IVE_FilterAndCscCtrl_t¶

Description

Defines template filter plus color space conversion composite function control information.

Defines

typedef struct MI_IVE_FilterAndCscCtrl_s
{
    MI_IVE_CscMode_e eMode; /*CSC working mode*/
    MI_S8 as8Mask[MI_IVE_MASK_SIZE_5X5]; /*Template parameter filter
    coefficient*/
    MI_U8 u8Norm; /*Normalization parameter, by right shift*/
}MI_IVE_FilterAndCscCtrl_t;

Member

Member Code Description

eMode Work mode.

as8Mask[MI_IVE_MASK_SIZE_5X5] 5x5 template coefficient, configured in row-major order.

u8Norm Normalization parameter. Parameter range: [0, 13].
Note
- Only 4 modes of YUV2RGB are supported.
Related Data Type and Interface

MI_IVE_CscMode_e

3.28. MI_IVE_SobelOutCtrl_e¶

Description

Defines Sobel Output control information.

Definition

typedef enum
{
    E_MI_IVE_SOBEL_OUT_CTRL_BOTH = 0x0, /*Output horizontal and
    vertical*/
    E_MI_IVE_SOBEL_OUT_CTRL_HOR  = 0x1, /*Output horizontal*/
    E_MI_IVE_SOBEL_OUT_CTRL_VER  = 0x2, /*Output vertical*/
    E_MI_IVE_SOBEL_OUT_CTRL_BUTT

}MI_IVE_SobelOutCtrl_e;

Member

Member Code	Description
E_MI_IVE_SOBEL_OUT_CTRL_BOTH	Output results of both template filtering and transposed template filtering.
E_MI_IVE_SOBEL_OUT_CTRL_HOR	Output template filtering results only.
E_MI_IVE_SOBEL_OUT_CTRL_VER	Output transposed template filtering results only.

Related Data Type and Interface

MI_IVE_SobelCtrl_t

3.29. MI_IVE_SobelCtrl_t¶

Description

Defines Sobel-like gradient calculation control information.

Definition

typedef struct MI_IVE_SobelCtrl_s
{
    MI_IVE_SobelOutCtrl_e eOutCtrl; /*Output format*/
    MI_S8 as8Mask[MI_IVE_MASK_SIZE_5X5]; /*Template parameter*/

}MI_IVE_SobelCtrl_t;

Member

Member Code Description

eOutCtrl Output control enumeration parameter.

as8Mask[MI_IVE_MASK_SIZE_5X5] 5x5 template coefficient, configured in row-major order.
Related Data Type and Interface

MI_IVE_SobelOutCtrl_e

3.30. MI_IVE_MagAndAngOutCtrl_e¶

Description

Defines gradient magnitude and angle calculation output format.

Definition

typedef enum
{
    E_MI_IVE_MAG_AND_ANG_OUT_CTRL_MAG         = 0x0,
    E_MI_IVE_MAG_AND_ANG_OUT_CTRL_MAG_AND_ANG = 0x1,
    E_MI_IVE_MAG_AND_ANG_OUT_CTRL_BUTT
}MI_IVE_MagAndAngOutCtrl_e;

Member

Member Code Description

E_MI_IVE_MAG_AND_ANG_OUT_CTRL_MAG Output magnitude only.

E_MI_IVE_MAG_AND_ANG_OUT_CTRL_MAG_AND_ANG Output both magnitude and angle.
Related Data Type and Interface

MI_IVE_MagAndAngCtrl_t

3.31. MI_IVE_MagAndAngCtrl_t¶

Description

Defines gradient magnitude and angle calculation control information.

Definition

typedef struct MI_IVE_MagAndAngCtrl_s
{
    MI_IVE_MagAndAngOutCtrl_e eOutCtrl;
    MI_U16 u16Thr;
    MI_S8 as8Mask[MI_IVE_MASK_SIZE_5X5]; /*Template parameter.*/
}MI_IVE_MagAndAngCtrl_t;

Member

Member Code Description

eOutCtrl Output format.

u16Thr Threshold value for magnitude thresholding.

as8Mask[MI_IVE_MASK_SIZE_5X5] 5x5 template coefficient, configured in row-major order.
Related Data Type and Interface

MI_IVE_MagAndAngOutCtrl_e

3.32. MI_IVE_DilateCtrl_t¶

Description

Defines Dilate control information.

Definition

typedef struct MI_IVE_DilateCtrl_s
{
    MI_U8 au8Mask[MI_IVE_MASK_SIZE_5X5]; /*The template parameter value
    must be 0 or 255.*/
}MI_IVE_DilateCtrl_t;

Member

Member Code Description

au8Mask[MI_IVE_MASK_SIZE_5X5] 5x5 template coefficient. Range: 0 or 255, configured in row-major order.

3.33. MI_IVE_ErodeCtrl_t¶

Description

Defines Erode control information.

Definition

typedef struct MI_IVE_ErodeCtrl_s
{
    MI_U8 au8Mask[MI_IVE_MASK_SIZE_5X5]; /*The template parameter value
    must be 0 or 255.*/
}MI_IVE_ErodeCtrl_t;

Member

Member Code Description

au8Mask[MI_IVE_MASK_SIZE_5X5] 5x5 template coefficient. Range: 0 or 255, configured in row-major order.

3.34. MI_IVE_ThreshMode_e¶

Description

Defines image thresholding output format.

Definition

typedef enum
{
    E_MI_IVE_THRESH_MODE_BINARY      = 0x0, /*srcVal <= lowThr, dstVal = minVal; srcVal > lowThr, dstVal = maxVal.*/
    E_MI_IVE_THRESH_MODE_TRUNC       = 0x1, /*srcVal <= lowThr, dstVal = srcVal; srcVal > lowThr, dstVal = maxVal.*/
    E_MI_IVE_THRESH_MODE_TO_MINVAL   = 0x2, /*srcVal <= lowThr, dstVal = minVal; srcVal > lowThr, dstVal = srcVal.*/
    E_MI_IVE_THRESH_MODE_MIN_MID_MAX = 0x3, /*srcVal <= lowThr, dstVal = minVal; lowThr < srcVal <= highThr, dstVal = midVal; srcVal > highThr, dstVal = maxVal.*/
    E_MI_IVE_THRESH_MODE_ORI_MID_MAX = 0x4, /*srcVal <= lowThr, dstVal = srcVal; lowThr < srcVal <= highThr, dstVal = midVal; srcVal > highThr, dstVal = maxVal.*/
    E_MI_IVE_THRESH_MODE_MIN_MID_ORI = 0x5, /*srcVal <= lowThr, dstVal = minVal; lowThr < srcVal <= highThr, dstVal = midVal; srcVal > highThr, dstVal = srcVal.*/
    E_MI_IVE_THRESH_MODE_MIN_ORI_MAX = 0x6, /*srcVal <= lowThr, dstVal = minVal; lowThr < srcVal <= highThr, dstVal = srcVal; srcVal > highThr, dstVal = maxVal.*/
    E_MI_IVE_THRESH_MODE_ORI_MID_ORI = 0x7, /*srcVal <= lowThr, dstVal = srcVal; lowThr < srcVal <= highThr, dstVal = midVal; srcVal > highThr, dstVal = srcVal.*/
    E_MI_IVE_THRESH_MODE_BUTT
}MI_IVE_ThreshMode_e;

Member

Member Code	Description
E_MI_IVE_THRESH_MODE_BINARY	srcVal ≤ lowThr, dstVal = minVal; srcVal > lowThr, dstVal = maxVal.
E_MI_IVE_THRESH_MODE_TRUNC	srcVal ≤ lowThr, dstVal = srcVal; srcVal > lowThr, dstVal = maxVal.
E_MI_IVE_THRESH_MODE_TO_MINVAL	srcVal ≤lowThr, dstVal = minVal; srcVal > lowThr, dstVal = srcVal.
E_MI_IVE_THRESH_MODE_MIN_MID_MAX	srcVal ≤ lowThr, dstVal = minVal; lowThr < srcVal ≤ highThr, dstVal = midVal; srcVal > highThr, dstVal = maxVal.
E_MI_IVE_THRESH_MODE_ORI_MID_MAX	srcVal ≤lowThr, dstVal = srcVal; lowThr < srcVal ≤ highThr, dstVal = midVal; srcVal > highThr, dstVal = maxVal.
E_MI_IVE_THRESH_MODE_MIN_MID_ORI	srcVal ≤lowThr, dstVal = minVal; lowThr < srcVal ≤highThr,dstVal = midVal; srcVal > highThr, dstVal = srcVal.
E_MI_IVE_THRESH_MODE_MIN_ORI_MAX	srcVal ≤ lowThr, dstVal = minVal; lowThr < srcVal ≤ highThr,dstVal = srcVal; srcVal > highThr, dstVal = maxVal.
E_MI_IVE_THRESH_MODE_ORI_MID_ORI	srcVal≤ lowThr, dstVal = srcVal; lowThr < srcVal ≤ highThr, dstVal = midVal; srcVal > highThr, dstVal = srcVal.

Note
- For calculation formula, please refer to the Note provided with respect to MI_IVE_Thresh.
Related Data Type and Interface

MI_IVE_ThreshCtrl_t

3.35. MI_IVE_ThreshCtrl_t¶

Description

Defines image thresholding control information.

Definition

typedef struct MI_IVE_ThreshCtrl_s
{
    MI_IVE_ThreshMode_e eMode;
    MI_U8 u8LowThr; /*user-defined threshold, 0<=u8LowThr<=255 */
    MI_U8 u8HighThr; /*user-defined threshold, if eMode<E_MI_IVE_THRESH_MODE_MIN_MID_MAX, u8HighThr is not used, else0<=u8LowThr<=u8HighThr<=255;*/
    MI_U8 u8MinVal; /*Minimum value when tri-level thresholding*/
    MI_U8 u8MidVal; /*Middle value when tri-level thresholding, if eMode<2, u32MidVal is not used; */
    MI_U8 u8MaxVal; /*Maxmum value when tri-level thresholding*/
}MI_IVE_ThreshCtrl_t;

Member

Member Code	Description
eMode	Thresholding operation mode.
u8LowThresh	Low threshold. Parameter range: [0,255].
u8HighThresh	High threshold. 0≤u8LowThresh≤u8HighThresh≤255.
u8MinVal	Min. value. Parameter range: [0,255].
u8MidVal	Median value. Parameter range: [0,255].
u8MaxVal	Max. value. Parameter range: [0,255].

Related Data Type and Interface

MI_IVE_ThreshMode_e

3.36. MI_IVE_SubMode_e¶

Description

Defines image subtraction output format.

Definition

typedef enum
{
    E_MI_IVE_SUB_MODE_ABS   = 0x0, /*Absolute value of the difference*/
    E_MI_IVE_SUB_MODE_SHIFT = 0x1, /*The output result is obtained by
    shifting the result one digit right to reserve the signed bit.*/
    E_MI_IVE_SUB_MODE_BUTT
}MI_IVE_SubMode_e;

Member

Member Code Description

E_MI_IVE_SUB_MODE_ABS Absolute value of the difference.

E_MI_IVE_SUB_MODE_SHIFT Output the result by right-shifting one bit, sign bit reserved.
Related Data Type and Interface

MI_IVE_SubCtrl_t

3.37. MI_IVE_SubCtrl_t¶

Description

Defines image subtraction control parameter.

Definition

typedef struct MI_IVE_SubCtrl_s
{
    MI_IVE_SubMode_e eMode;
}MI_IVE_SubCtrl_t;

Member

Member Code Description

eMode Image subtraction mode.
Related Data Type and Interface

MI_IVE_SubMode_e

3.38. MI_IVE_IntegOutCtrl_e¶

Description

Defines integral map output control parameter.

Definition

typedef enum
{
    E_MI_IVE_INTEG_OUT_CTRL_COMBINE = 0x0,
    E_MI_IVE_INTEG_OUT_CTRL_SUM     = 0x1,
    E_MI_IVE_INTEG_OUT_CTRL_SQSUM   = 0x2,
    E_MI_IVE_INTEG_OUT_CTRL_BUTT
}MI_IVE_IntegOutCtrl_e;

Member

Member Code	Description
E_MI_IVE_INTEG_OUT_CTRL_COMBINE	Output integral map sum and sum of square.
E_MI_IVE_INTEG_OUT_CTRL_SUM	Output integral map sum only.
E_MI_IVE_INTEG_OUT_CTRL_SQSUM	Output integral map sum of square only.

Related Data Type and Interface

MI_IVE_IntegCtrl_t

3.39. MI_IVE_IntegCtrl_t¶

Description

Defines integral map calculation control parameter.

Definition

typedef struct MI_IVE_IntegCtrl_s
{
    MI_IVE_IntegOutCtrl_e eOutCtrl;
}MI_IVE_IntegCtrl_t;

Member

Member Code Description

eOutCtrl Integral map output control parameter.
Related Data Type and Interface

MI_IVE_IntegOutCtrl_e

3.40. MI_IVE_ThreshS16Mode_e¶

Description

Defines 16-bit signed image thresholding mode.

Definition

typedef enum
{
    E_MI_IVE_THRESH_S16_MODE_S16_TO_S8_MIN_MID_MAX = 0x0,
    E_MI_IVE_THRESH_S16_MODE_S16_TO_S8_MIN_ORI_MAX = 0x1,
    E_MI_IVE_THRESH_S16_MODE_S16_TO_U8_MIN_MID_MAX = 0x2,
    E_MI_IVE_THRESH_S16_MODE_S16_TO_U8_MIN_ORI_MAX = 0x3,
    E_MI_IVE_THRESH_S16_MODE_BUTT
}MI_IVE_ThreshS16Mode_e;

Member

Member Code	Description
E_MI_IVE_THRESH_S16_MODE_S16_ TO_S8_MIN_MID_MAX	srcVal ≤ lowThr, dstVal = minVal; lowThr < srcVal ≤highThr, dstVal = idVal; srcVal > highThr, dstVal = maxVal;
E_MI_IVE_THRESH_S16_MODE_S16_ TO_S8_MIN_ORI_MAX	srcVal ≤ lowThr, dstVal = minVal; lowThr < srcVal ≤highThr, dstVal = rcVal; srcVal > highThr, dstVal = maxVal;
E_MI_IVE_THRESH_S16_MODE_S16_ TO_U8_MIN_MID_MAX	srcVal ≤ lowThr, dstVal = minVal; lowThr < srcVal ≤highThr, dstVal = idVal; srcVal > highThr, dstVal = maxVal;
E_MI_IVE_THRESH_S16_MODE_S16_ TO_U8_MIN_ORI_MAX	srcVal ≤ lowThr, dstVal = minVal; lowThr < srcVal ≤highThr, dstVal = rcVal; srcVal > highThr, dstVal = maxVal;

Note
- For calculation formula, please refer to the Note provided with respect to MI_IVE_ThreshS16.
Related Data Type and Interface

MI_IVE_ThreshS16Ctrl_t

3.41. MI_IVE_ThreshS16Ctrl_t¶

Description

Defines 16-bit signed image thresholding control parameter.

Definition

typedef struct MI_IVE_ThreshS16Ctrl_s
{
    MI_IVE_ThreshS16Mode_e eMode;
    MI_S16 s16LowThr; /*user-defined threshold*/
    MI_S16 s16HighThr; /*user-defined threshold*/
    MI_IVE_Length8bit_u un8MinVal; /*Minimum value when tri-level thresholding*/
    MI_IVE_Length8bit_u un8MidVal; /*Middle value when tri-level thresholding*/
    MI_IVE_Length8bit_u un8MaxVal; /*Maxmum value when tri-level thresholding*/
}MI_IVE_ThreshS16Ctrl_t;

Member

Member Code Description

eMode Thresholding operation mode.

s16LowThr Low threshold.

s16HighThr High threshold.

un8MinVal Min. value.

un8MidVal Median value.

un8MaxVal Max. value.
Note
- For calculation formula, please refer to the Note provided with respect to MI_IVE_ThreshS16.
Related Data Type and Interface

MI_IVE_ThreshS16Mode_e

3.42. MI_IVE_ThreshU16Mode_e¶

Description

Defines 16-bit unsigned image thresholding mode.

Definition

typedef enum
{
    E_MI_IVE_THRESH_U16_MODE_U16_TO_U8_MIN_MID_MAX = 0x0,
    E_MI_IVE_THRESH_U16_MODE_U16_TO_U8_MIN_ORI_MAX = 0x1,
    E_MI_IVE_THRESH_U16_MODE_BUTT
}MI_IVE_ThreshU16Mode_e;

Member

Member Code	Description
E_MI_IVE_THRESH_U16_MODE_U16_ TO_U8_MIN_MID_MAX	srcVal ≤ lowThr, dstVal = minVal; lowThr < srcVal ≤highThr, dstVal = midVal; srcVal > highThr, dstVal = maxVal;
E_MI_IVE_THRESH_U16_MODE_U16_ TO_U8_MIN_ORI_MAX	srcVal ≤ lowThr, dstVal = minVal; lowThr < srcVal ≤highThr, dstVal = srcVal; srcVal > highThr, dstVal = maxVal;

Note
- For calculation formula, please refer to the Note provided with respect to MI_IVE_ThreshU16.
Related Data Type and Interface

MI_IVE_ThreshU16Ctrl_t

3.43. MI_IVE_ThreshU16Ctrl_t¶

Description

Defines 16-bit unsigned image thresholding control parameter.

Definition

typedef struct MI_IVE_ThreshU16Ctrl_s
{
    MI_IVE_ThreshU16Mode_e eMode;
    MI_U16 u16LowThr;
    MI_U16 u16HighThr;
    MI_U8 u8MinVal;
    MI_U8 u8MidVal;
    MI_U8 u8MaxVal;
}MI_IVE_ThreshU16Ctrl_t;

Member

Member Code	Description
eMode	Thresholding operation mode.
u16LowThr	Low threshold.
u16HighThr	High threshold.
u8MinVal	Min. value. Parameter range: [0,255].
u8MidVal	Median value. Parameter range: [0,255].
u8MaxVal	Max. value. Parameter range: [0,255].

Note
- For calculation formula, please refer to the Note provided with respect to MI_IVE_ThreshU16.
Related Data Type and Interface

MI_IVE_ThreshU16Mode_e

3.44. MI_IVE_16BitTo8BitMode_e¶

Description

Defines 16-bit image data to 8-bit image data conversion mode.

Definition

typedef enum
{
    E_MI_IVE_16BIT_TO_8BIT_MODE_S16_TO_S8      = 0x0,
    E_MI_IVE_16BIT_TO_8BIT_MODE_S16_TO_U8_ABS  = 0x1,
    E_MI_IVE_16BIT_TO_8BIT_MODE_S16_TO_U8_BIAS = 0x2,
    E_MI_IVE_16BIT_TO_8BIT_MODE_U16_TO_U8      = 0x3,
    E_MI_IVE_16BIT_TO_8BIT_MODE_BUTT
}MI_IVE_16BitTo8BitMode_e;

Member

Member Code	Description
E_MI_IVE_16BIT_TO_8BIT_MODE_S16_TO_S8	S16 data to S8 data linear conversion.
E_MI_IVE_16BIT_TO_8BIT_MODE_S16_TO_U8_ABS	Linearly convert S16 data to S8 data and takes the absolute value to get U8 data.
E_MI_IVE_16BIT_TO_8BIT_MODE_S16_TO_U8_BIAS	Linearly convert S16 data to S8 data and truncate to U8 data after translation.
E_MI_IVE_16BIT_TO_8BIT_MODE_U16_TO_U8	S16 data to U8 data linear conversion.

Note
- For calculation formula, please refer to Note in MI_IVE_ThreshU16.
Related Data Type and Interface

MI_IVE_16bitTo8BitCtrl_t

3.45. MI_IVE_16bitTo8BitCtrl_t¶

Description

Defines 16-bit image data to 8-bit image data conversion control parameter.

Definition

typedef struct MI_IVE_16bitTo8BitCtrl_s
{
    MI_IVE_16BitTo8BitMode_e eMode;
    MI_U16 u16Denominator;
    MI_U8 u8Numerator;
    MI_S8 s8Bias;
}MI_IVE_16bitTo8BitCtrl_t;

Member

Member Code	Description
eMode	16-bit data to 8-bit data conversion mode.
u16Denominator	Denominator of linear conversion. Parameter range: [max
u8Numerator	Numerator of linear conversion. Parameter range: [0,255].
s8Bias	Translated item of linear conversion. Parameter range: [-128,127].

Note
- For calculation formula, please refer to Note in MI_IVE_ThreshU16.
- u8Numerator ≤u16Denominator, and u16Denominator≠0
Related Data Type and Interface

MI_IVE_16BitTo8BitMode_e

3.46. MI_IVE_OrdStatFilterMode_e¶

Description

Defines sequential statistic filtering mode.

Definition

typedef enum
{
    E_MI_IVE_ORD_STAT_FILTER_MODE_MEDIAN = 0x0,
    E_MI_IVE_ORD_STAT_FILTER_MODE_MAX    = 0x1,
    E_MI_IVE_ORD_STAT_FILTER_MODE_MIN    = 0x2,
    E_MI_IVE_ORD_STAT_FILTER_MODE_BUTT
}MI_IVE_OrdStatFilterMode_e;

Member

Member Code	Description
E_MI_IVE_ORD_STAT_FILTER_MODE_MEDIAN	Median filter.
E_MI_IVE_ORD_STAT_FILTER_MODE_MAX	Max. filter value, equivalent to the dilation of grayscale.
E_MI_IVE_ORD_STAT_FILTER_MODE_MIN	Min. filter value, equivalent to the erosion of grayscale.

Related Data Type and Interface

MI_IVE_OrdStatFilter_t

3.47. MI_IVE_OrdStatFilter_t¶

Description

Defines sequential statistic filtering control parameter.

Definition

typedef struct MI_IVE_OrdStatFilter_s
{
    MI_IVE_OrdStatFilterMode_e eMode;
}MI_IVE_OrdStatFilter_t;

Member

Member Code Description

eMode Sequential statistic filtering mode.
Related Data Type and Interface

MI_IVE_OrdStatFilterMode_e

3.48. MI_IVE_MapLutMem_t¶

Description

Defines map operator lookup table memory information.

Definition

typedef struct MI_IVE_MapLutMem_s
{
    MI_U8 au8Map[MI_IVE_MAP_NUM];
}MI_IVE_MapLutMem_t;

Member

Member Code Description

au8Map[MI_IVE_MAP_NUM] Map lookup table array.

3.49. MI_IVE_EqualizeHistCtrlMem_t¶

Description

Defines histogram equalization auxiliary memory.

Definition

typedef struct MI_IVE_EqualizeHistCtrlMem_s
{
    MI_U32 au32Hist[MI_IVE_HIST_NUM];
    MI_U8 au8Map[MI_IVE_MAP_NUM];
}MI_IVE_EqualizeHistCtrlMem_t;

Member

Member Code Description

au32Hist[MI_IVE_HIST_NUM] Histogram statistics output.

au8Map[MI_IVE_MAP_NUM] Map lookup table gained by statistical histogram calculation.
Related Data Type and Interface

MI_IVE_EqualizeHistCtrl_t

3.50. MI_IVE_EqualizeHistCtrl_t¶

Description

Defines histogram equalization control parameter.

Definition

typedef struct MI_IVE_EqualizeHistCtrl_s
{
    MI_IVE_MemInfo_t stMem;
}MI_IVE_EqualizeHistCtrl_t;

Member

Member Code Description

stMem The memory should at least have the size of (MI_IVE_EqualizeHistCtrlMem_t).
Related Data Type and Interface

MI_IVE_EqualizeHistCtrlMem_t

3.51. MI_IVE_AddMode_e¶

Description

Define the add calculation mode.

Definition

typedef enum
{
    E_MI_IVE_ADD_MODE_ROUNDING = 0x0,
    E_MI_IVE_ADD_MOD_CLIPPING  = 0x1,
    E_MI_IVE_ADD_MODE_MAX
}MI_IVE_AddMode_e;

Member

Member Code Description

E_MI_IVE_ADD_MODE_ROUNDING Rounding mode

E_MI_IVE_ADD_MOD_CLIPPING Clipping mode
Related Data Type and Interface

MI_IVE_AddCtrl_t

3.52. MI_IVE_AddCtrl_t¶

Description

Defines image weighted addition control parameter.

Definition

typedef struct MI_IVE_AddCtrl_s
{
    MI_IVE_AddMode_e eMode;
    MI_U0Q16 u0q16X; /*x of "xA+yB"*/
    MI_U0Q16 u0q16Y; /*y of "xA+yB"*/
}MI_IVE_AddCtrl_t;

Member

Member Code	Description
eMode	Calculation mode.
u0q16X	Weighted addition by the weight “x” in “xA+yB.” Parameter range: [1, 65535].
u0q16Y	Weighted addition by the weight “y” in “xA+yB.” Parameter range:

Related Data Type and Interface

MI_IVE_AddMode_e

3.53. MI_IVE_NccDstMem_t¶

Description

Defines NCC output memory information.

Definition

typedef struct MI_IVE_NccDstMem_s
{
    MI_U64 u64Numerator;
    MI_U64 u64QuadSum1;
    MI_U64 u64QuadSum2;
}MI_IVE_NccDstMem_t;

Member

Member Code	Description
u64Numerator	Numerator of NCC calculation formula: $<span><span class="MathJax_Preview">\sum^w_{i=1}\sum^h_{j=1}(I_{src1}(i,j)I_{src2}(i,j))</span><script type="math/tex">\sum^w_{i=1}\sum^h_{j=1}(I_{src1}(i,j)I_{src2}(i,j))$
u64QuadSum1	Denominator, the inner part of the root number, of NCC calculation formula: $<span><span class="MathJax_Preview">\sum^w_{i=1}\sum^h_{j=1}(I^2_{src1}(i,j))</span><script type="math/tex">\sum^w_{i=1}\sum^h_{j=1}(I^2_{src1}(i,j))$
u64QuadSum2	Denominator, the inner part of the root number, of NCC calculation formula: $<span><span class="MathJax_Preview">\sum^w_{i=1}\sum^h_{j=1}(I^2_{src2}(i,j))</span><script type="math/tex">\sum^w_{i=1}\sum^h_{j=1}(I^2_{src2}(i,j))$

Note
- For calculation formula, please refer to the Note provided with respect to MI_IVE_Ncc.

3.54. MI_IVE_Region_t¶

Description

Defines connected region information.

Definition

typedef struct MI_IVE_Region_s
{
    MI_U32 u32Area; /*Represented by the pixel number*/
    MI_U16 u16Left; /*Circumscribed rectangle left border*/
    MI_U16 u16Right; /*Circumscribed rectangle right border*/
    MI_U16 u16Top; /*Circumscribed rectangle top border*/
    MI_U16 u16Bottom; /*Circumscribed rectangle bottom border*/
}MI_IVE_Region_t;

Member

Member Code	Description
u32Area	Area size of connected region, represented by number of pixels of the connected region.
u16Left	Leftmost coordinate of the circumscribed rectangle of connected region.
u16Right	Rightmost coordinate of the circumscribed rectangle of connected region.
u16Top	Uppermost coordinate of the circumscribed rectangle of connected region.
u16Bottom	Lowermost coordinate of the circumscribed rectangle of connected region.

Related Data Type and Interface

MI_IVE_CcBlob_t

3.55. MI_IVE_CcBlob_t¶

Description

Defines connected region label output information.

Definition

typedef struct MI_IVE_CcBlob_s
{
    MI_U16 u16CurAreaThr; /*Threshold of the result regions' area*/
    MI_S8 s8LabelStatus; /*-1: Labeled failed ; 0: Labeled successfully*/
    MI_U8 u8RegionNum; /*Number of valid region, non-continuous stored*/
    MI_IVE_Region_t astRegion[MI_IVE_MAX_REGION_NUM];
}MI_IVE_CcBlob_t;

Member

Member Code	Description
u16CurAreaThr	The regions in astRegion with area size smaller than this threshold will be deleted and merged when doing region deletion.
s8LabelStatus	Connected region labelled successfully or not. -1: Labelling failed, overflow is occurred; 0: Labelling successful.
u8RegionNum	Effective number of connected regions.
astRegion[MI_IVE_MAX_REGION_NUM]	Connected region information.

Chips/Platforms	u16CurAreaThr	s8LabelStatus	Regions Information with Label Index 1~254	Excluded Regions Information after Merging Deleted Regions
Souffle	0	-1: Labelling failed, overflow occurred; 0: Labelling successful	astRegion[1]~astRegion[254]	astRegion[0]
Others	Determined by deletion process	-1: Labelling failed(u8RegionNum == 0) or Labelling successful(u8RegionNum != 0) with deletion process. 0: Labelling successful	astRegion[0]~astRegion[253]	astRegion[254]

Note
- CcBlob→u8RegionNum indicates the effective number of connected regions, and the maximum number of effective region is 254.
- u16CurAreaThr will be outputed after region deletion when the effective number of connected regions is over 254, for details of deletion, please refer to MI_IVE_CclCtrl_t.
Related Data Type and Interface

MI_IVE_Region_t

MI_IVE_CclCtrl_t

3.56. MI_IVE_CclMode_e¶

Description

Defines connected region mode.

Definition

typedef enum
{
    E_MI_IVE_CCL_MODE_8C = 0x0, /*8-connectivity*/
    E_MI_IVE_CCL_MODE_4C = 0x1, /*4-connectivity*/
    E_MI_IVE_CCL_MODE_BUTT
}MI_IVE_CclMode_e;

Member

Member Code Description

E_MI_IVE_CCL_MODE_4C 4-connectivity.

E_MI_IVE_CCL_MODE_8C 8-connectivity.
Note
- Only Souffle series supports this enumeration, others use 8-connectivity mode as default.

3.57. MI_IVE_CclCtrl_t¶

Description

Defines connected region label control parameter.

Definition

typedef struct MI_IVE_CclCtrl_s
{
    MI_U16 u16InitAreaThr;
    MI_U16 u16Step;
    MI_IVE_CclMode_e eCclMode;
}MI_IVE_CclCtrl_t;

Member

Member Code	Description
u16InitAreaThr	Initial area threshold. Parameter range: [0, 65535]. Reference value: 4.
u16Step	Area threshold growth step. Parameter range: [1,65535]. Reference value: 2.
eCclMode	The mode of CCL. Only Souffle series support

Note
- When the number of connected regions are over 254, use pstCclCtrl→u16InitAreaThr to exclude the region with small area. If the number of connected regions are still over 254 after region deletion, the threshold of area will be increased by step pstCclCtrl→u16Step and stored in u16CurAreaThe, and do deletion until the number of connected regions is less than 254 or the threshold of area reaches the maximum value 65535.
Related Data Type and Interface

MI_IVE_CcBlob_t

3.58. MI_IVE_GmmCtrl_t¶

Description

Defines GMM background modelling control parameter.

Definition

typedef struct MI_IVE_GmmCtrl_s
{
    MI_U22Q10 u22q10NoiseVar; /*Initial noise Variance*/
    MI_U22Q10 u22q10MaxVar; /*Max Variance*/
    MI_U22Q10 u22q10MinVar; /*Min Variance*/
    MI_U0Q16 u0q16LearnRate; /*Learning rate*/
    MI_U0Q16 u0q16BgRatio; /*Background ratio*/
    MI_U8Q8 u8q8VarThr; /*Variance Threshold*/
    MI_U0Q16 u0q16InitWeight; /*Initial Weight*/
    MI_U8 u8ModelNum; /*Model number: 3 or 5*/
}MI_IVE_GmmCtrl_t;

Member

Member Code	Description
u22q10NoiseVar	Initial noise variance. Parameter range: [0x1, 0xFFFFFF]. For grayscale GMM, please refer to noiseSigma * noiseSigma of OpenCV MOG grayscale model. Reference value: 1515(1<<10). For RGB GMM, please refer to noiseSigma * noiseSigma of OpenCV MOG RGB model 3 . Reference value: 31515(1<<10). NOTE: Souffle series do not need to set this parameter.
u22q10MaxVar	Maximum value of model variance. Parameter range; [0x1, 0xFFFFFF]. Please refer to fVarMax of OpenCV MOG2. Reference value: 3*4000<<10 (RGB), 2000<<10 (grayscale).
u22q10MinVar	Minimum value of model variance. Parameter range: [0x1, 22q10MaxVar]. Please refer to fVarMin of OpenCV MOG2. Reference value: 600<<10 (RGB), 200<<10 (grayscale).
u0q16LearnRate	Learning rate. Parameter range: [1, 65535]. Please refer to learningRate of OpenCV MOG2. Reference value: if (frameNum<500) (1/frameNum)((1<<16)-1); else ((1/500)((1<<16)-1).
u0q16BgRatio	Background ratio threshold. Parameter range: [1, 65535]. Please refer to backgroundRatio of OpenCV MOG. Reference value: 0.8*((1<<16)-1).
u8q8VarThr	Variance threshold. Parameter range: [1, 65535]. Please refer to varThreshold of OpenCV MOG. This parameter is used to determine if a pixel hits the current model. Reference value: 6.25*(1<<8).
u0q16InitWeight	Initial weight. Parameter range: [1, 65535]. Please refer to defaultInitialWeight of OpenCV MOG. Reference value: 0.05((1<<16)-1). NOTE: Souffle series do not need to set this parameter.*
u8ModelNum	Model number. Parameter range:

3.59. MI_IVE_CannyStackSize_t¶

Description

Defines strong edge point stack size structure in the first half of Canny edge calculation.

Definition

typedef struct MI_IVE_CannyStackSize_s
{
    MI_U32 u32StackSize; /*Stack size for output*/
    MI_U8 u8Reserved[MI_IVE_CANNY_STACK_RESERVED_SIZE]; /*For 16 byte align*/
}MI_IVE_CannyStackSize_t;

Member

Member Code Description

u32StackSize Stack size (number of strong edge points).

u8Reserved[MI_IVE_CANNY_STACK_RESERVED_SIZE] Reserved.

3.60. MI_IVE_CannyHysEdgeCtrl_t¶

Description

Defines control parameter in Canny edge first-half calculation task.

Definition

typedef struct MI_IVE_CannyHysEdgeCtrl_s
{
    MI_IVE_MemInfo_t stMem;
    MI_U16 u16LowThr;
    MI_U16 u16HighThr;
    MI_S8 as8Mask[MI_IVE_MASK_SIZE_5X5];
} MI_IVE_CannyHysEdgeCtrl_t;

Member

Member Code	Description
stMem	Auxiliary memory. For details on the memory allocation and size, please refer to the Note provided with respect to MI_IVE_CannyHysEdge.
u16LowThr	Low threshold. Parameter range: [0,255].
u16HighThr	High threshold. Parameter range: [u16LowThr,255].
as8Mask[MI_IVE_MASK_SIZE_5X5]	Parameter template used for gradient calculation.

3.61. MI_IVE_LbpCmpMode_e¶

Description

Defines LBP calculation comparison mode.

Definition

typedef enum
{
    E_MI_IVE_LBP_CMP_MODE_NORMAL  = 0x0,
    /* P(x)-P(center) >= un8BitThr.s8Val, s(x)=1; else s(x)=0; */
    E_MI_IVE_LBP_CMP_MODE_ABS     = 0x1,
    /* abs(P(x)-P(center)) >= un8BitThr.u8Val, s(x)=1; else s(x)=0; */
    E_MI_IVE_LBP_CMP_MODE_ABS_MUL = 0x2
    E_MI_IVE_LBP_CMP_MODE_MAX
}MI_IVE_LbpCmpMode_e;

Member

Member Code	Description
E_MI_IVE_LBP_CMP_MODE_NORMAL	LBP normal comparison mode.
E_MI_IVE_LBP_CMP_MODE_ABS	LBP absolute comparison mode.
E_MI_IVE_LBP_CMP_MODE_ABS_MUL	LBP absolute multiply comparison mode

Note
- For calculation formula, please refer to the Note provided with respect to MI_IVE_Lbp.
Related Data Type and Interface

MI_IVE_LbpCtrrl_t

3.62. MI_IVE_LbpChalMode_e¶

Description

Defines LBP input channel mode.

Definition

typedef enum
{
    E_MI_IVE_LBP_CHAL_MODE_U8C1 = 0x0,
    E_MI_IVE_LBP_CHAL_MODE_U8C2 = 0x1,
    E_MI_IVE_LBP_CHAL_MODE_MAX
} MI_IVE_LbpChalMode_e;

Member

Member Code Description

E_MI_IVE_LBP_CHAL_MODE_U8C1 Only one input

E_MI_IVE_LBP_CHAL_MODE_U8C2 Two input
Related Data Type and Interface

MI_IVE_LbpCtrrl_t

3.63. MI_IVE_LbpCtrrl_t¶

Description

Defines LBP texture calculation control parameter.

Definition

typedef struct MI_IVE_LbpCtrrl_s
{
    MI_IVE_LbpCmpMode_e eMode;
    MI_IVE_LbpChalMode_e chMode;
    MI_IVE_Length8bit_u un8BitThr;
}MI_IVE_LbpCtrrl_t;

Member

Member Code	Description
eMode	LBP comparison mode.
chMode	LBP channel mode
un8BitThr	LBP comparison threshold. Parameter range for E_MI_IVE_LBP_CMP_MODE_NORMAL: [-128, 127]. Parameter range for E_MI_IVE_LBP_CMP_MODE_ABS: [0, 255]. Parameter range for E_MI_IVE_LBP_CMP_MODE_ABS: [0, 7].

Note
- For calculation formula, please refer to the Note provided with respect to MI_IVE_Lbp.
Related Data Type and Interface

MI_IVE_LbpCmpMode_e

MI_IVE_LbpChalMode_e

MI_IVE_Length8bit_u

3.64. MI_IVE_NormGradOutCtrl_e¶

Description

Defines enumeration type of normalized gradient information calculation task output control.

Definition

typedef enum
{
    E_MI_IVE_NORM_GRAD_OUT_CTRL_HOR_AND_VER = 0x0,
    E_MI_IVE_NORM_GRAD_OUT_CTRL_HOR         = 0x1,
    E_MI_IVE_NORM_GRAD_OUT_CTRL_VER         = 0x2,
    E_MI_IVE_NORM_GRAD_OUT_CTRL_COMBINE     = 0x3,
    E_MI_IVE_NORM_GRAD_OUT_CTRL_BUTT
}MI_IVE_NormGradOutCtrl_e;

Member

Member Code	Description
E_MI_IVE_NORM_GRAD_OUT_CTRL_ HOR_AND_VER	Output H and V component maps of gradient information simultaneously. (For H and V definition, please refer to Parameter of MI_ IVE_NormGrad.)
E_MI_IVE_NORM_GRAD_OUT_CTRL_ HOR	Output H component map of gradient information.
E_MI_IVE_NORM_GRAD_OUT_CTRL_ VER	Output V component map of gradient information.
E_MI_IVE_NORM_GRAD_OUT_CTRL_ COMBINE	Output H, V component map in package format of gradient information.

Related Data Type and Interface

MI_IVE_NormGradCtrl_t

3.65. MI_IVE_NormGradCtrl_t¶

Description

Defines normalized gradient information calculation control parameter.

Definition

typedef struct MI_IVE_NormGradCtrl_s
{
    MI_IVE_NormGradOutCtrl_e eOutCtrl;
    MI_S8 as8Mask[MI_IVE_MASK_SIZE_5X5];
    MI_U8 u8Norm;
}MI_IVE_NormGradCtrl_t;

Member

Member Code	Description
eOutCtrl	Gradient information output control mode.
as8Mask[MI_IVE_MASK_SIZE_5X5]	Template required of gradient calculation.
u8Norm	Normalization parameter. Parameter range: [1, 13].

Related Data Type and Interface

MI_IVE_NormGradOutCtrl_e

3.66. MI_IVE_MvS9Q7_t¶

Description

Defines LK optical flow displacement structure.

Description

typedef struct MI_IVE_MvS9Q7_s
{
    MI_S32 s32Status; /*Result of tracking: 0-success; -1-failure*/
    MI_S9Q7 s9q7Dx; /*X-direction component of the movement*/
    MI_S9Q7 s9q7Dy; /*Y-direction component of the movement*/
}MI_IVE_MvS9Q7_t;

Member

Member Code Description

s32Status Feature point tracking result. 0: Successful. 1: Failed.

s9q7Dx Displacement of X-direction component.

s9q7Dy Displacement of Y-direction component.

3.67. MI_IVE_LkOpticalFlowCtrl_t¶

Description

Defines LK optical flow calculation control parameter.

Definition

typedef struct MI_IVE_LkOpticalFlowCtrl_s
{
    MI_U16 u16CornerNum; /*Number of the feature points,<200*/
    MI_U0Q8 u0q8MinEigThr; /*Minimum eigenvalue threshold*/
    MI_U8 u8IterCount; /*Maximum iteration times*/
    MI_U0Q8 u0q8Epsilon; /*Threshold of iteration for dx^2 + dy^2 < u0q8Epsilon */
}MI_IVE_LkOpticalFlowCtrl_t;

Member

Member Code	Description
u16CornerNum	Number of input corner points / feature points. Parameter range: [1,200].
u0q8MinEigThr	Minimum Eigen value threshold. Parameter range: [1,255].
u8IterCount	Maximum number of iterations. Parameter range: [1,20].
u0q8Epsilon	Eteration convergence condition: dx^2 + dy^2 < u0q8Epsilon. Parameter range: [1,255]. Reference value: 2.

3.68. MI_IVE_SadMode_e¶

Description

Defines SAD calculation mode.

Definition

typedef enum
{
    E_MI_IVE_SAD_MODE_MB_4X4   = 0x0, /*4x4*/
    E_MI_IVE_SAD_MODE_MB_8X8   = 0x1, /*8x8*/
    E_MI_IVE_SAD_MODE_MB_16X16 = 0x2, /*16x16*/
    E_MI_IVE_SAD_MODE_BUTT
}MI_IVE_SadMode_e;

Member

Member Code	Description
E_MI_IVE_SAD_MODE_MB_4X4	Calculation of SAD by 4x4 pixel block.
E_MI_IVE_SAD_MODE_MB_8X8	Calculation of SAD by 8x8 pixel block.
E_MI_IVE_SAD_MODE_MB_16X16	Calculation of SAD by 16x16 pixel block.

Related Data Type and Interface

MI_IVE_SadCtrl_t

3.69. MI_IVE_SadOutCtrl_e¶

Description

Defines SAD output control mode.

Definition

typedef enum
{
    E_MI_IVE_SAD_OUT_CTRL_16BIT_BOTH = 0x0, /*Output 16 bit sad and thresh*/
    E_MI_IVE_SAD_OUT_CTRL_8BIT_BOTH  = 0x1, /*Output 8 bit sad and thresh*/
    E_MI_IVE_SAD_OUT_CTRL_16BIT_SAD  = 0x2, /*Output 16 bit sad*/
    E_MI_IVE_SAD_OUT_CTRL_8BIT_SAD   = 0x3, /*Output 8 bit sad*/
    E_MI_IVE_SAD_OUT_CTRL_THRESH     = 0x4, /*Output thresh,16 bits sad */
    E_MI_IVE_SAD_OUT_CTRL_BUTT
}MI_IVE_SadOutCtrl_e;

Member

Member Code	Description
E_MI_IVE_SAD_OUT_CTRL_16BIT_BOTH	16-bit SAD map and thresholding map output mode.
E_MI_IVE_SAD_OUT_CTRL_8BIT_BOTH	8-bit SAD map and thresholding map output mode.
E_MI_IVE_SAD_OUT_CTRL_16BIT_SAD	16-bit SAD map output mode.
E_MI_IVE_SAD_OUT_CTRL_8BIT_SAD	8-bit SAD map output mode.
E_MI_IVE_SAD_OUT_CTRL_THRESH	Thresholding map output mode.

Related Data Type and Interface

MI_IVE_SadCtrl_t

3.70. MI_IVE_SadCtrl_t¶

Description

Defines SAD control parameter.

Definition

typedef struct MI_IVE_SadCtrl_s
{
    MI_IVE_SadMode_e eMode;
    MI_IVE_SadOutCtrl_e eOutCtrl;
    MI_U16 u16Thr; /*srcVal <= u16Thr, dstVal = minVal;srcVal > u16Thr, dstVal = maxVal.*/
    MI_U8 u8MinVal; /*Min value*/
    MI_U8 u8MaxVal; /*Max value*/
}MI_IVE_SadCtrl_t;

Member

Member Code	Description
eMode	SAD calculation mode.
eOutCtrl	SAD output control mode.
u16Thr	Value for thresholding against calculated SAD map.
u8MinVal	Value for thresholding lower than u16Thr.
u8MaxVal	Value for thresholding higher than u16Thr.

Related Data Type and Interface

MI_IVE_SadMode_e

MI_IVE_SadOutCtrl_e

3.71. MI_IVE_BernsenCtrl_t¶

Description

Defines Bernsen thresh control parameters.

Definition

typedef struct MVE_IVE_BernsenCtrl_s
{
    MI_IVE_BernsenMode_e enMode;
    MI_U8 u8WinSize;
    MI_U8 u8MaxVal;
} MVE_IVE_BernsenCtrl_t;

Member

Member Code	Description
eMode	Bernsen thresh mode.
u8WinSize	Window size for the local threshold calculation. Value range:
u8MaxVal	Thresh in MVE_BERNSEN_MODE_THRESH mode in which the global threshold is involved. Value range: [0, 255]

Related Data Type and Interface

MI_IVE_BernsenMode_e

3.72. MI_IVE_BernsenMode_e¶

Description

Defines the Bernsen thresh mode.

Definition

typedef enum
{
    E_MI_IVE_BERNSEN_MODE_NORMAL = 0x0,
    E_MI_IVE_BERNSEN_MODE_THRESH = 0x1,
    E_MI_IVE_BERNSEN_MODE_MAX
} MVE_IVE_BernsenMode_e;

Member

Member Code	Description
E_MI_IVE_BERNSEN_MODE_NORMAL	Simple Bernsen thresh
E_MI_IVE_BERNSEN_MODE_THRESH	Thresh based on the global threshold and local Bernsen threshold
E_MI_IVE_BERNSEN_MODE_MAX	Error mode.

Related Data Type and Interface

MI_IVE_BernsenCtrl_t

3.73. MI_IVE_LineFilterHorCtrl_t¶

Description

Defines control parameters for filtering the horizontal density of binary images.

Definition

typedef struct MVE_IVE_LineFilterHorCtrl_s
{
    MI_U8 u8GapMinLen;
    MI_U8 u8DensityThr;
    MI_U8 u8HorThr;
} MVE_IVE_LineFilterHorCtrl_t;

Member

Member Code	Description
u8GapMinLen	Minimum black line length. For details, see thr1 in the note field of MI_IVE_LineFilterHor. Value range: [1, 20] Reference value: 10
u8DensityThr	Density threshold. For details, see thr2 in the Note field of MI_IVE_LineFilterHor. Value range: [1, 50] Reference value: 20
u8HorThr	Horizontal line length threshold. For details, see thr3 in the note field of MI_IVE_LineFilterHor. Value range: [1, 50] Reference value: 20

3.74. MI_IVE_LineFilterVerCtrl_t¶

Description

Defines control parameters for filtering the vertical density of binary images.

Definition

typedef struct MVE_IVE_LineFilterVerCtrl_s
{
    MI_U8 u8VerThr;
} MVE_IVE_LineFilterVerCtrl_t _t;

Member

Member Code Description

u8VerThr Vertical line length threshold. For details, see thr in the note field of MI_IVE_LineFilterVer. Value range: [1, 64] Reference value: 30

3.75. MI_IVE_NoiseRemoveHor_t¶

Description

Defines the horizontal noise removal control parameter for the binary image.

Definition

typedef struct MVE_IVE_NoiseRemoveHorCtrl_s
{
    MI_U8 u8HorThr;
    MI_U8 u8HorThrMax;
} MVE_IVE_NoiseRemoveHorCtrl_t;

Member

Member Code	Description
u8HorThr	Length threshold for determining horizontal noises, see thr1 in the note field of MI_IVE_NoiseRemoveHor. Value range: [1, 100] Reference value: 25
u8HorThrMax	Maximum length threshold for determining horizontal , see thr2 in the note field of MI_IVE_NoiseRemoveHor. Value range: [1, 100] Reference value: 79

Note
- u8HorThrMax is always bigger than u8HorThr.

3.76. MI_IVE_NoiseRemoveVer_t¶

Description

Defines the vertical noise removal control parameter for the binary image.

Definition

typedef struct MVE_IVE_NoiseRemoveVerCtrl_s
{
    MI_U8 u8VerThr;
    MI_U8 u8VerThrMax;
} MVE_IVE_NoiseRemoveVerCtrl_t;

Member

Member Code	Description
u8VerThr	Length threshold for determining vertical noises, see thr1 in the note field of MI_IVE_NoiseRemoveVer. Value range: [1, 100] Reference value: 25
u8VerThrMax	Maximum Length threshold for determining vertical noises, see thr2 in the note field of MI_IVE_NoiseRemoveVer. Value range: [1, 100] Reference value: 79

Note
- u8VerThrMax is always bigger than u8VerThr.

3.77. MI_IVE_AdpThreshCtrl_t¶

Description

Defines adaptive thresh control parameters.

Definition

typedef struct MVE_IVE_AdpThreshCtrl_s
{
    MI_U8 u8RateThr;
    MI_U8 u8HalfMaskx;
    MI_U8 u8HalfMasky;
    MI_U8 s16Offset;
    MI_U8 u8ValueThr;
} MVE_IVE_AdpThreshCtrl_t;

Member

Member Code	Description
u8RateThr	Threshold rate for determining adaptive thresh. For details, see RateThr in the formula of MI_IVE_AdpThresh. Value range: [1, 20] Default: 10
u8HalfMaskx	Half of Mask size width for determining adaptive thresh. For details, see w in the formula of MI_IVE_AdpThresh. Value range: [1, 40] Default: 10
u8HalfMasky	Half of Mask size height for determining adaptive thresh. For details, see h in the formula of MI_IVE_AdpThresh. Value range: [1, 40] Default: 35
s16Offset	Offset for determining adaptive thresh. For details, see Offset in the formula of MI_IVE_AdpThresh. Value range: [-128, 127] Default: -100
u8ValueThr	Threshold pixel value for determining adaptive thresh. For details, see ValueThr in the formula of MI_IVE_AdpThresh. Value range: [1, 255] Default: 100

3.78. MI_IVE_ResizeMode_e¶

Description

Defines the resize control parameters.

Definition

typedef enum
{
    E_MI_IVE_RESIZE_TYPE_U8C1         = 0x0,
    E_MI_IVE_RESIZE_TYPE_U8C3_PLANAR  = 0x1,
    E_MI_IVE_RESIZE_TYPE_U8C3_PACKAGE = 0x2,
    E_MI_IVE_RESIZE_TYPE_YUV420SP     = 0x3,
    E_MI_IVE_RESIZE_TYPE_MAX
} MVE_IVE_ResizeMode_e;

Member

Member Code	Description
E_MI_IVE_RESIZE_TYPE_U8C1	Single-channel image of which each pixel is expressed by an 8-bit unsigned data segment.
E_MI_IVE_RESIZE_TYPE_U8C3_PLANAR	Three-channel image (stored in planar format) of which each pixel is expressed by three 8-bit unsigned data segments.
E_MI_IVE_RESIZE_TYPE_U8C3_PACKAGE	Three-channel image (stored in package format) of which each pixel is expressed by three 8-bit unsigned data segments.
E_MI_IVE_RESIZE_TYPE_YUV420SP	YUV420 semi-planar image.
E_MI_IVE_RESIZE_TYPE_MAX	Error mode.

Related Data Type and Interface

MI_IVE_ResizeCtrl_t

3.79. MI_IVE_ResizeMethod_e¶

Description

Defines the resize method.

Definition

typedef enum
{
    E_MI_IVE_RESIZE_METHOD_BILINEAR = 0x0,
    E_MI_IVE_RESIZE_METHOD_AREA     = 0x1,
    E_MI_IVE_RESIZE_METHOD_MAX
} MI_IVE_ResizeMethod_e;

Member

Member Code	Description
E_MI_IVE_RESIZE_METHOD_BILINEAR	Use Bilinear method to do resize, scaling up and down are supported.
E_MI_IVE_RESIZE_METHOD_AREA	Use area method to do resize, only scaling down is supported.
E_MI_IVE_RESIZE_TYPE_MAX	Error mode.

Note
- Only Souffle series supports this enumeration, others use Bilinear mode as default.
Related Data Type and Interface

MI_IVE_ResizeCtrl_t

3.80. MI_IVE_ResizeCtrl_t¶

Description

Defines the resize control parameters.

Definition

typedef struct _MVE_IVE_ResizeCtrl_s
{
    MVE_IVE_ResizeMode_e enMode;
    MI_IVE_ResizeMethod_e eResizeMethod;
} MVE_IVE_ResizeCtrl_t;

Member

Member Code Description

enMode Input image mode. For detail, see MI_IVE_ResizeMode_e

eResizeMethod Resize method.
Only Souffle series support
Related Data Type and Interface

MI_IVE_ResizeMode_e

MI_IVE_ResizeMethod_e

3.81. MI_IVE_BatCtrl_t¶

Description

Defines the bat control parameters.

Definition

typedef struct _MVE_IVE_BatCtrl_s
{
    MVE_IVE_BatMode_e enMode;
    MI_U16 _t u16HorTimes;
    MI_U16 _t u16VerTimes;
} MVE_IVE_BatCtrl_t;

Member

Member Code Description

enMode Input image mode. For detail, see MI_IVE_BatMode_e

u16HorTimes Threshold of horizontal direction.

u16VerTimes Threshold of vertical direction.
Related Data Type and Interface

MI_IVE_BatMode_e

3.82. MI_IVE_BatMode_e¶

Description

Define the output control mode.

Definition

typedef enum
{
    E_MI_IVE_BAT_OUT_CTRL_BOTH = 0x0,
    E_MI_IVE_BAT_OUT_CTRL_HOR  = 0x1,
    E_MI_IVE_BAT_OUT_CTRL_VER  = 0x2,
    E_MI_IVE_BAT_OUT_CTRL_MAX
} MVE_IVE_BatMode_e

Member

Member Code	Description
E_MI_IVE_BAT_OUT_CTRL_BOTH	Horizontal and vertical mode.
E_MI_IVE_BAT_OUT_CTRL_HOR	Horizontal mode.
E_MI_IVE_BAT_OUT_CTRL_VER	Vertical mode
E_MI_IVE_BAT_OUT_CTRL_MAX	Error mode.

Related Data Type and Interface

MI_IVE_BatCtrl_t

3.83. MI_IVE_AccCtrl_t¶

Description

Define the acc control parameter.

Definition

typedef struct MVE_IVE_AccCtrl_s
{
    MVE_IVE_AccMode_e enMode;
} MVE_IVE_AccCtrl_t;

Member

Member Code Description

enMode Input image mode. For detail, see MI_IVE_AccMode_e
Related Data Type and Interface

MI_IVE_AccMode_e

3.84. MI_IVE_AccMode_e¶

Description

Define the acc input control mode.

Definition

typedef enum
{
    E_MI_IVE_ACC_MODE_INCREASE            = 0x0,
    E_MI_IVE_ACC_MODE_DECREASE            = 0x1,
    E_MI_IVE_ACC_MODE_INCREASE_MAP_255TO1 = 0x2,
    E_MI_IVE_ACC_MODE_MAX
} MVE_IVE_AccMode_e

Member

Member Code	Description
E_MI_IVE_ACC_MODE_INCREASE	Increase mode.
E_MI_IVE_ACC_MODE_DECREASE	Decrease mode.
E_MI_IVE_ACC_MODE_INCREASE_MAP_255TO1	Increase 255 to 1 mode.
E_MI_IVE_ACC_MODE_MAX	Error mode.

Related Data Type and Interface

MI_IVE_AccCtrl_t

3.85. MI_IVE_MatrTranfMode_e¶

Description

Define the input channel mode of matrix transform.

Definition

typedef enum
{
    E_MI_IVE_MATRIX_TRANSFORM_TYPE_C1 = 0x0,
    E_MI_IVE_MATRIX_TRANSFORM_TYPE_C2 = 0x1,
    E_MI_IVE_MATRIX_TRANSFORM_TYPE_C3 = 0x2,
    E_MI_IVE_MATRIX_TRANSFORM_TYPE_MAX
}MVE_IVE_MatrTranfMode_e;

Member

Member Code	Description
E_MI_IVE_MATRIX_TRANSFORM_TYPE_C1	Only use one input source.
E_MI_IVE_MATRIX_TRANSFORM_TYPE_C2	Two input sources.
E_MI_IVE_MATRIX_TRANSFORM_TYPE_C3	Thress input sources.

Related Data Type and Interface

MI_IVE_MatrTranfCtrl_t

3.86. MI_IVE_MatrTranfCtrlMode_e¶

Description

Define the input control mode of matrix transform.

Definition

typedef enum
{
    E_MI_IVE_MATRIX_TRANSFORM_CTRL_ROUNDING = 0x0,
    E_MI_IVE_MATRIX_TRANSFORM_CTRL_CLIPPING = 0x1,
    E_MI_IVE_MATRIX_TRANSFORM_CTRL_MAX
} MVE_IVE_MatrTranfCtrlMode_e;

Member

Member Code Description

E_MI_IVE_MATRIX_TRANSFORM_CTRL_ROUNDING Rounding mode.

E_MI_IVE_MATRIX_TRANSFORM_CTRL_CLIPPING Clipping mode.

E_MI_IVE_MATRIX_TRANSFORM_CTRL_MAX Error mode.
Related Data Type and Interface

MI_IVE_MatrTranfCtrl_t

3.87. MI_IVE_MatrTranfCtrl_t¶

Description

Define the control parameters of matrix transform.

Definition

typedef struct MI_IVE_MatrTranfCtrl_S
{
    MVE_IVE_MatrTranfMode_e enMode; *Input channel mode*/
    MVE_IVE_MatrTranfCtrlMode_e CtrlMode;
    MI_S32 s32MatrixArray[9]; //Official
} MI_IVE_MatrTranfCtrl_t;

Member

Member Code	Description
enMode	Input image mode. For detail, see MI_IVE_MatrTranfMode_e.
CtrlMode	Input controling mode.
s32MatrixArray[9];	Cofficients of matrix, configured in row-major order.

Note
- If the enMode = E_MI_IVE_MATRIX_TRANSFORM_TYPE_C1, you only need to set value s32MatrixArray[0].
- If the enMode = E_MI_IVE_MATRIX_TRANSFORM_TYPE_C2, the value s32MatrixArray[0] ~ s32MatrixArray[3] must be set.
- If the enMode = E_MI_IVE_MATRIX_TRANSFORM_TYPE_C3, the all matrix value should be set.
Related Data Type and Interface

MI_IVE_MatrTranfMode_e

MI_IVE_MatrTranfCtrlMode_e

3.88. MI_IVE_SHIFT_DETECT_MODE_e¶

Description

Define the shift detector input control mode.

Definition

typedef enum
{
    E_MI_IVE_SHIFT_DETECT_MODE_SINGLE = 0x0,
    E_MI_IVE_SHIFT_DETECT_MODE_MULTI  = 0x1,
    E_MI_IVE_SHIFT_DETECT_MODE_MAX
}MVE_IVE_MatrTranfMode_e;

Member

Member Code Description

E_MI_IVE_SHIFT_DETECT_MODE_SINGLE Only use one input source.

E_MI_IVE_SHIFT_DETECT_MODE_MULTI Two input sources.

E_MI_IVE_SHIFT_DETECT_MODE_MAX Error mode
Related Data Type and Interface

MI_IVE_SHIFT_DETECT_CTRL_t

3.89. MI_IVE_SHIFT_DETECT_CTRL_t¶

Description

Define control parameter of the shift detector.

Definition

typedef struct MVE_IVE_SHIFT_DETECT_CTRL_S
{
    MVE_IVE_SHIFT_DETECT_MODE_e enMode;
    MI_U8 pyramid_level;
    MI_U8 search_range;
    MI_U16 u16Left;
    MI_U16 u16Top;
    MI_U16 u16Width;
    MI_U16 u16Height;
} MI_IVE_SHIFT_DETECT_CTRL_t;

Member

Member Code	Description
enMode	Input image mode. For detail, see MI_IVE_MatrTranfMode_e
pyramid_level	The size of the image pyramid.
search_range	The searching region around the bounding box
u16Left	The top left x location of the bounding box
u16Top	The top left y location of the bounding box
u16Width	The width of the bounding box
u16Height	The height of the bounding box

Note
- If enMode = E_MI_IVE_SHIFT_DETECT_MODE_SINGLE, u16Width and u16Height represent the width and height of target region.
- If enMode = E_MI_IVE_SHIFT_DETECT_MODE_MULTI, u16Width and u16Height represent square size of global tracking.
Related Data Type and Interface

MI_IVE_SHIFT_DETECT_MODE_e

3.90. MI_IVE_BgBlurMode_e¶

Description

Define the BGBlur mode.

Definition

typedef enum
{
    E_MI_IVE_BGBLUR_MODE_BLUR            = 0x0,
    E_MI_IVE_BGBLUR_MODE_REPLACE         = 0x1,
    E_MI_IVE_BGBLUR_MODE_MOSAIC          = 0x2,
    E_MI_IVE_BGBLUR_MODE_BLUR_AND_MOSAIC = 0x3,
    E_MI_IVE_BGBLUR_MODE_MAX
} MI_IVE_BgBlurMode_e;

Member

Member Code	Description
E_MI_IVE_BGBLUR_MODE_BLUR	The background blur mode of BGBlur.
E_MI_IVE_BGBLUR_MODE_REPLACE	The background replacing mode of BGBlur. Only Souffle series support
E_MI_IVE_BGBLUR_MODE_MOSAIC	The background mosaic mode of BGBlur. Only iFord series support
E_MI_IVE_BGBLUR_MODE_BLUR_AND_MOSAIC	The background blur and mosaic mode of BGBlur. Only iFord series support
E_MI_IVE_BGBLUR_MODE_MAX	Error mode

Related Data Type and Interface

MI_IVE_BgBlurCtrl_t

3.91. MI_IVE_BgBlurMaskOp_e¶

Description

Define the Mask Preprocess mode of BGBlur.

Definition

typedef enum
{
    E_MI_IVE_BGBLUR_MASK_OP_DILATE = 0x0,
    E_MI_IVE_BGBLUR_MASK_OP_NONE   = 0x1,
    E_MI_IVE_BGBLUR_MASK_OP_ERODE  = 0x2,
    E_MI_IVE_BGBLUR_MASK_OP_MAX
} MI_IVE_BgBlurMaskOp_e;

Member

Member Code	Description
E_MI_IVE_BGBLUR_MASK_OP_DILATE	The Mask dilation preprocess mode
E_MI_IVE_BGBLUR_MASK_OP_NONE	No Mask preprocess mode
E_MI_IVE_BGBLUR_MASK_OP_ERODE	The Mask corrosion preprocess mode
E_MI_IVE_BGBLUR_MASK_OP_MAX	Error mode

Note
- The output of BGBlur will arise halo/flicker in foreground or penetration in background when the Mask content is not fitted with the shape of real foreground, in this case, use Mask preprocess can improve the output of BGBlur.
- E_MI_IVE_BGBLUR_MASK_OP_DILATE is recommended when background has less texture and the contrast between background and foreground is biggish.
- E_MI_IVE_BGBLUR_MASK_OP_NONE or E_MI_IVE_BGBLUR_MASK_OP_ERODE is recommended when background has rich texture.
Related Data Type and Interface

MI_IVE_BgBlurCtrl_t

3.92. MI_IVE_BgBlurCtrl_t¶

Description

Define the control parameter of BGBlur.

Definition

typedef struct MI_IVE_BgBlurCtrl_S
{
    MI_IVE_BGBLUR_MODE_e eBgBlurMode;
    MI_U8 u8MaskThr;
    MI_U8 u8BlurLv;
    MI_U8 u8ScalingStage;
    MI_IVE_BgBlurMaskOp_e eBgBlurMaskOp;
    MI_U8 u8SaturationLv;
    MI_U8 u8MosaicSize;
} MI_IVE_BgBlurCtrl_t;

Member

Member Code	Description
eBgBlurMode	The BGBlur mode.
u8MaskThr	The threshold for binarization of Mask source image, parameter range: [0, 255].
u8BlurLv	The level of background blur, parameter range: [0, 255]. Only valid in Souffle series
u8ScalingStage	The scaling stage of background blur, parameter range: [1, 15].
eBgBlurMaskOp	The Mask Preprocess mode of BGBlur.
u8SaturationLv	The saturation level of U/V component in output image, parameter range: [0, 128]. Only iFord series support
u8MosaicSize	The granularity of background mosaic, parameter range: [2, 4, 6, 8, 10]. Only iFord series support

Note
- Parameter u8BlurLv is used for indicating the weight of Alpha blending, the strength of blur reaches maximum when u8BlurLv is 0, other description can be referred in MI_IVE_AlphaBlending.
- Parameter u8ScalingStage is used for indicating scaling stage of small original image. When configuring background blur, the small original image will be resized, and the blur degree of the acquired background is directly proportional to the scaling stage, the strength reaches maximum when u8ScalingStage is 15.
- Parameter u8SaturationLv is used for indicating the saturation level of U/V component in output image, background is gray image when u8SaturationLv is 0, full color image when u8SaturationLv is 128.
- Parameter u8MosaicSize is used for indicating the granularity of background mosaic, the mosaic strength reaches maximum when u8MosaicSize is 10.
Related Data Type and Interface

MI_IVE_BgBlurMode_e

MI_IVE_BgBlurMaskOp_e

4. ERROR CODE¶

The following table lists the Intellegient Acceleration Engine API ERROR CODE.

Table 3-1: Intellegient Acceleration Engine API Error Codes

Error Code	Macro Definition	Description
0xA0002001	MI_IVE_ERR_INVALID_DEVID	Invalid ID
0xA0002002	MI_IVE_ERR_INVALID_CHNID	Invalid channel ID or area handle
0xA0002003	MI_IVE_ERR_ILLEGAL_PARAM	Illegal parameter
0xA0002004	MI_IVE_ERR_EXIST	Device, channel or resource already exists
0xA0002005	MI_IVE_ERR_UNEXIST	Device, channel or resource to be used or destroyed does not exist
0xA0002006	MI_IVE_ERR_NULL_PTR	Null pointer found in function parameter
0xA0002007	MI_IVE_ERR_NOT_CONFIG	Module not configured
0xA0002008	MI_IVE_ERR_NOT_SUPPORT	Parameter or function not supported
0xA0002009	MI_IVE_ERR_NOT_PERM	Operation, e.g. attempt to modify static configuration parameter, not permitted
0xA000200C	MI_IVE_ERR_NOMEM	Memory allocaton failed, e.g. system memory not enough
0xA000200D	MI_IVE_ERR_NOBUF	Buffer allocation failed, e.g. image buffer area too broad
0xA000200E	MI_IVE_ERR_BUF_EMPTY	No image in buffer area
0xA000200F	MI_IVE_ERR_BUF_FULL	Image full in buffer area
0xA0002010	MI_IVE_ERR_SYS_NOTREADY	System not initialized or corresponding module not loaded
0xA0002011	MI_IVE_ERR_BADADDR	Invalid address
0xA0002012	MI_IVE_ERR_BUSY	System busy
0xA0002013	MI_IVE_ERR_CHN_NOT_STARTED	Channel not start
0xA0002014	MI_IVE_ERR_CHN_NOT_STOPED	Channel not stop
0xA0002015	MI_IVE_ERR_NOT_INIT	Module not initialized before being used
0xA0002019	MI_IVE_ERR_SYS_TIMEOUT	System timeout
0xA000201F	MI_IVE_ERR_FAILED	Unexpected error

5. PROCFS INTRODUCTION¶

5.1. cat¶

Debug info analysis
```
# cat /proc/mi_modules/mi_ive/mi_ive0
```
Debug info analysis

Record the current debug info of IVE device which can be used for debugging or testing, and in which device info, request list management info, timeout info and channels info are included.

Parameters Description

Parameters		Description
Dev info	DevState	Device state
	ClkRate	Used clock rate
	IrqNum	Interrupt number
	TaskCnt	Unfinished task count
	FireCnt	Processed task count
	IsrDoneCnt	Finished task count
	IsrBackPresCnt	Back pressure interrupt count
	IsrOtherCnt	Abnormal interrupt count
	ResetCnt	Reset count
Request List info	No.	Request list index
	1stTaskState	The First task state in Request list N
	ReadyTaskCnt	The count of task with state Ready in Request list N
	DoingTaskCnt	The count of task with state Doing in Request list N
	DoneTaskCnt	The count of task with state Done in Request list N
	InQTaskCnt	The count of task with state InQ in Request list N
Timeout info	No.	Timeout info idx
	Op	Recorded operator when timeout
	Pri	Recorded task priority when timeout
	inW inH Fmt	Recorded input width/height/format when timeout
	OutW OutH Fmt	Recorded output width/height/format when timeout
	OpConfig	Other recorded configuration when timeout
Chns info	ChnId	Channel identifier
	CurOp	Current used operator
	CurWidth	Current input width
	CurHeight	Current input height
	OpHistory	Used histroy of operator
	HwCallTimes	Times of using operator with hardware acceleration
	SwCallTimes	Times of using operator without hardware acceleration
	HwSwitchSwTimes	Times of using operator with parameters which is not satisfied with hardware acceleration requirement

Return Value	Description
Handle index	Range: [0, MI_IVE_HANDLE_MAX)
MI_IVE_HANDLE_MAX	Failed.

Member Code	Description
phyPhyAddr	One-dimensional data physical address.
pu8VirAddr	One-dimensional data virtual address.
u32Size	One-dimensional data byte size.

Member Code	Description
s25q7X	x coordinate of point, in SQ25.7 representation.
s25q7Y	y coordinate of point, in SQ25.7 representation.

Member Code	Description
eMode	Work mode.
as8Mask[MI_IVE_MASK_SIZE_5X5]	5x5 template coefficient, configured in row-major order.
u8Norm	Normalization parameter. Parameter range: [0, 13].

Member Code	Description
eOutCtrl	Output control enumeration parameter.
as8Mask[MI_IVE_MASK_SIZE_5X5]	5x5 template coefficient, configured in row-major order.

Member Code	Description
E_MI_IVE_MAG_AND_ANG_OUT_CTRL_MAG	Output magnitude only.
E_MI_IVE_MAG_AND_ANG_OUT_CTRL_MAG_AND_ANG	Output both magnitude and angle.

Member Code	Description
eOutCtrl	Output format.
u16Thr	Threshold value for magnitude thresholding.
as8Mask[MI_IVE_MASK_SIZE_5X5]	5x5 template coefficient, configured in row-major order.

Member Code	Description
E_MI_IVE_SUB_MODE_ABS	Absolute value of the difference.
E_MI_IVE_SUB_MODE_SHIFT	Output the result by right-shifting one bit, sign bit reserved.

Member Code	Description
eMode	Thresholding operation mode.
s16LowThr	Low threshold.
s16HighThr	High threshold.
un8MinVal	Min. value.
un8MidVal	Median value.
un8MaxVal	Max. value.

Member Code	Description
au32Hist[MI_IVE_HIST_NUM]	Histogram statistics output.
au8Map[MI_IVE_MAP_NUM]	Map lookup table gained by statistical histogram calculation.

Member Code	Description
E_MI_IVE_ADD_MODE_ROUNDING	Rounding mode
E_MI_IVE_ADD_MOD_CLIPPING	Clipping mode

Member Code	Description
E_MI_IVE_CCL_MODE_4C	4-connectivity.
E_MI_IVE_CCL_MODE_8C	8-connectivity.

Member Code	Description
u32StackSize	Stack size (number of strong edge points).
u8Reserved[MI_IVE_CANNY_STACK_RESERVED_SIZE]	Reserved.

Member Code	Description
E_MI_IVE_LBP_CHAL_MODE_U8C1	Only one input
E_MI_IVE_LBP_CHAL_MODE_U8C2	Two input

Member Code	Description
s32Status	Feature point tracking result. 0: Successful. 1: Failed.
s9q7Dx	Displacement of X-direction component.
s9q7Dy	Displacement of Y-direction component.

Member Code	Description
enMode	Input image mode. For detail, see MI_IVE_ResizeMode_e
eResizeMethod	Resize method. Only Souffle series support

Member Code	Description
enMode	Input image mode. For detail, see MI_IVE_BatMode_e
u16HorTimes	Threshold of horizontal direction.
u16VerTimes	Threshold of vertical direction.

Member Code	Description
E_MI_IVE_MATRIX_TRANSFORM_CTRL_ROUNDING	Rounding mode.
E_MI_IVE_MATRIX_TRANSFORM_CTRL_CLIPPING	Clipping mode.
E_MI_IVE_MATRIX_TRANSFORM_CTRL_MAX	Error mode.

Member Code	Description
E_MI_IVE_SHIFT_DETECT_MODE_SINGLE	Only use one input source.
E_MI_IVE_SHIFT_DETECT_MODE_MULTI	Two input sources.
E_MI_IVE_SHIFT_DETECT_MODE_MAX	Error mode