An Efficient Matching Algorithm for Segment-Based Stereo Vision Using Dynamic Programming Technique

摘要

An efficient matching method for segment-based stereo vision is proposed. A potential matching graph which describes the connectivity between candidate matching pairs of segments is built. Establishing correspondence is then reduced to a problem of searching for the optimal path that maximizes a similarity measure. The optimal path is found efficiently without the adverse effects of combinatorial explosion by using a dynamic programming technique. The validity of th emethod is confirmed by experiments with actual images.