Dense disparity estimation with a divide-and-conquer disparity space image technique

摘要: A new divide-and-conquer technique for disparity estimation is proposed in this paper. This performs feature matching following the high confidence first principle, starting with strongest point stereo pair of scanlines. Once established, ordering constraint allows original intra-scanline problem to be divided into two smaller subproblems. Each subproblem can then solved recursively until there no reliable within subintervals. very efficient dense map images rich features. For general scenes, paired up disparity-space image (DSI) compute maps integrated occlusion detection. In approach, part algorithm handles stronger features and DSI-based pixels between points detection occlusions. An extension standard also presented compliment algorithm. Experiments demonstrate effectiveness DSI