A Region-Based Stereo Algorithm

Abstract

Feature- point- based stereo and segment-based stereo have been intensively studied. But why not region-based stereo? Region is more characteristic than point and segment. Matching between regions is easier and faster, if region segmentation is successful. Unfortunately, however, it is the failure to satisfy this precondition that has been the obstacle. Region segmentation in practice is subtle and unstable, and it is not guaranteed that region boundaries in stereo images correspond to same physical positions. Our approach to tackling this problem is to sharpen intensity contrast between regions by projecting a random pattern to the visual field. The proposed algorithm has been implemented to match both computer-generated random pattern stereo images and pattern-projected real images. Results show that the algorithm is robust, fast and tolerant of small vertical disparities.