Function Description

Description

The following table describes the functions of the vision preprocessing core (VPC) and the function support of each version. All APIs described in this chapter are asynchronous. The API call delivers a task rather than executes a task. After this API is called, call the synchronization API (for example, aclrtSynchronizeStream) to ensure that the task is complete.

Function	Description	Atlas 200/300/500 Inference Product	Atlas Training Series Product
Cropping	Crops one or more selected regions of interest (ROIs) from one or more images.	√	√
Resizing:	8K resizing and non-8K resizing are supported for images based on resolutions. Resizes an ROI (supporting uncompressed format) or multiple ROIs from a single image (supporting uncompressed format). Other resizing modes, such as resizing an image without cropping it and resizing an image with the aspect ratio preserved.	√	√
Pasting	Crops a selected ROI from the input image, resizes the cropped image, and loads it to the output buffer. The output buffer may be a blank image (when the allocated output buffer is empty) or an existing canvas (when an image has been read into the allocated output buffer). Note that the pasting concept here refers only to the case when the output buffer is an existing canvas.	√	√
Stitching	Crops selected ROIs from the input image, resizes the cropped images, and loads them to specified areas in the output buffer.	√	√
Image pyramid	Successively downsamples an image using a Gaussian filter.	x	x
Histogram analysis	Collects statistics on the pixel value distribution of each image channel (RGB/YUV).	x	x
Remapping	Remaps an image to a new one based on the given configuration.	x	x
Border making	Creates a border around an image (also referred to as padding).	x	x
Format conversion	Converts between RGB and YUV.	√	√
Image grayscale	Converts a color image into a grayscale image. Note that a grayscale input image produces only a grayscale output image.	√ Only the data of the Y component is extracted from the YUV420SP output image.	√ Only the data of the Y component is extracted from the YUV420SP output image.

Function Diagrams (Atlas 200/300/500 Inference Product) (Atlas Training Series Product)

Figure 1 VPC function diagram (cropping + resizing + pasting)

Figure 2 VPC function diagram (stitching)

Figure 3 Resizing with the aspect ratio preserved (paste ROI centered on the output image)

Figure 4 Resizing with the aspect ratio preserved (paste ROI vertically centered on the output image)

Figure 5 Resizing with the aspect ratio preserved (paste ROI horizontally centered on the output image)

Function Description

Description

Function Diagrams (Atlas 200/300/500 Inference Product) (Atlas Training Series Product)

See Also