A device for processing disparity estimation image

Abstract

The embodiment of the invention provides a device for processing a disparity estimation image. The device comprises a verification module for circularly receiving a first pixel point and a second pixel point; the first pixel point received at first serves as a target pixel point in a circulating mode, left-right consistency verification is conducted on the basis of the second pixel point, and thetarget pixel point is marked according to the verification result; the marked target pixel points are output to a cavity filling module; the cavity filling module fills the target pixel point markedas the cavity pixel point according to the target pixel point marked as the stable pixel point; the target pixel point is output to a fusion module; the sub-pixel refinement module performs sub-pixelrefinement processing on the first pixel point according to a preset sub-pixel refinement method, and outputs the first pixel point as a target pixel point to the fusion module; and the fusion modulefuses the target pixel point received from the cavity filling module with the target pixel point received from the sub-pixel refinement module. Therefore, the processing efficiency of the pixel pointsin the disparity estimation value image is improved.

Description

technical field [0001] The present invention relates to the technical field of image processing, in particular to a device for processing parallax estimation images. Background technique [0002] Stereo matching (Stereo Matching) is: the process of matching two images as a stereo matching pair, that is, estimating each pixel in one of the images (for example: left view), and the pixel in the other image (for example, right view) The corresponding stereo matching (Stereo Matching) relationship. Wherein, the above two images as a stereo matching pair may be: two images collected for the same scene by image acquisition devices at different viewpoints, and the above image acquisition devices at different viewpoints may be binocular cameras. [0003] In the process of performing stereo matching on two images as a stereo matching pair, the corresponding value of each pixel in the image can be calculated based on the cost value of each image at different preset parallax levels and...

AI Technical Summary

Problems solved by technology

[0007] In the prior art, it is a serial processing of one pixel point by one pixel point, that is, after a pixel point is completely processed, the next pixel point is processed to obtain the final parallax value of all pixels, and the serial processing The process takes a long time and is not efficient

Images