Parallax refining algorithm based on matching cost updating and image segmentation

Abstract

The invention discloses a parallax refining algorithm based on matching cost updating and image segmentation. The method includes the steps of conducting image segmentation for a reference image IR and a target image IT, using a three-dimensional matching algorithm based on window aggregation, conducting initial three-dimensional matching for the reference image IR and the target image IT to obtain the initial parallax graph DAR of the reference image and the initial parallax graph of the target image, calculating the updated reference image matching cost, obtaining the aggregation weight omega R (iR, jR) of the reference image and the aggregation weight omega T (iT, jT) of the target image, obtaining the parallax graph DBR of the reference image and the parallax graph DBT of the target image, and obtaining the final dense parallax graph DR of the reference image and the dense parallax graph DT of the target image.

Description

technical field [0001] The invention relates to stereo matching in the field of computer stereo vision, and relates to a parallax refinement algorithm based on matching cost update and image segmentation, which can be used for three-dimensional reconstruction of images and provide guidance in medical images and media equipment. Background technique [0002] Stereo vision is a research hotspot in the field of computer vision, and its purpose is to obtain three-dimensional information between scene images. This technology has been widely used in many fields, such as stereoscopic display, scene reconstruction, pedestrian detection, etc. A complete stereo vision system includes: camera calibration, epipolar correction, stereo matching, and 3D reconstruction. [0003] Stereo matching is the most critical technology in stereo vision. Its main function is to obtain matching information between scene images. It mainly consists of four parts: matching cost calculation, matching cost...

AI Technical Summary

Problems solved by technology

In order to obtain a high-precision stereo matching disparity map, the method based on the weighted median filter is introduced into the disparity refining stage, but this type of algorithm usually needs to specify the size of the disparity refining window, and the pixel points outside the window are not considered in the refining process. The influence of parallax refinement makes the accuracy of parallax refinement limited

[0004] In practical applications, the traditional parallax refinement algorithm based on weighted median filtering has certain limitations, and it is difficult to obtain high stereo matching accuracy. This is also a problem faced by many researchers in stereo matching

Images