Dense stereo correspondence based on recursive adaptive size multi-windowing
Sylvia Oi Yee Chan, Ya-Ping Wong, Jacob K. Daniel · Siti Hasmah Digital Library-MMU Institutiona Repository (Multimedia University) · 2003
An efficient technique of stereo matching to compute dense disparity map is presented. The technique is point-oriented and uses multiple windows to enhance the strategy of finding a best match. In addit ion to that, a novel technique for adapting the window size based on multiple windows is discussed. The window size is recursively adapted using normalized sum of squared differences similarity measure and also the output of the multi-windowing method. Several constraints like epipolar constraint, uniqueness constraint and mutual correspondence constraint are also incorporated into the algorithm to prune bad matches. The results of the technique which has been tested on real images are also presented.