September 2019
Spotlight Summary by Rui Wang
Stereo matching based on multi-scale fusion and multi-type support regions
This article addresses a challenging task in stereo matching, that is matching pixels in regions with a low amount or free of texture. Stereo matching or disparity estimation is the process of finding correspondences of pixels in stereoscopic views as well as computing the amount of the displacement (disparity) of corresponding pixels. Human depth perception permanently and unconsciously solves this problem in the brain. However, in the "computer version," it is a problem addressed for a long time, but still unsolved. One of the main reasons is that in the absence of texture, correspondences of pixels become highly ambiguous. Li and coauthors have designed a complex algorithm with novel ideas to tackle the problem. First, color and gradient features of pixels are considered and fused for a more precise determination of differences of pixels. Second, when it is no longer reliable to only consider local features, stereo matching is processed in larger (up to global) scales. Finally, regions are categorized and treated differently in disparity refinement to provide a better estimation of disparity. Given the importance of this problem, the progress made in this article not only will help people to get a better understanding of human depth perception, but it also will broaden the utilization of computer vision into additional actual applications.
You must log in to add comments.
Add Comment
You must log in to add comments.
Article Information
Stereo matching based on multi-scale fusion and multi-type support regions
Haibin Li, Yakun Gao, Ziyue Huang, and Yakun Zhang
J. Opt. Soc. Am. A 36(9) 1523-1533 (2019) View: Abstract | HTML | PDF