Key points are not available for this paper at this time.
A computational framework is introduced for matching a pair of stereo images which, in contrast to existing algorithms, features a self-contained local matching module cascaded with a global matching module. Local matching outputs a 3-D grey-scale image in which each and every point has an intensity measuring the goodness of a possible match. Global matching reduces to surface detection in this image. To detect the surface, it is first enhanced, employing a hyperpyramid data structure. Unlike traditional multiresolution approaches, which are based on the coarse-to-fine continuation method, the authors' multilevel method emphasizes a fine-to-coarse process in which local support is accumulated. The algorithm is concise, efficient and above all, gives good results for complex scenes.>
Yang et al. (Mon,) studied this question.