基于控制点的分层双向动态规划立体匹配算法

提出一种基于控制点的分层双向动态规划立体匹配算法．首先，利用改进Volumetric迭代算法获取具有高可靠度的控制点，将其作为具有正确视差的匹配点．其次，在高可靠度控制点的指导下，利用分层双向动态规划算法在DSI（disparity-space image）视差空间图中进行初匹配，进而在Delta DSI（delta disparity-space image）视差变化空间图中进行精匹配，从而获取高密度视差图．实验结果表明，该算法不仅可以改善传统直接动态规划立体匹配算法产生的带状条纹瑕疵，而且计算速度较快，匹配结果也优于传统动态规划的匹配结果．

A Control-point-based Stereo Matching Algorithm for Layered and Bidirectional Dynamic Programming

A stereo matching algorithm based on control points is proposed for layered and bidirectional dynamic programming.Firstly,the improved volumetric iteration algorithm is utilized to acquire the high-reliability control points,which are used as the matching points with right disparity.Secondly,under the guidance of the high-reliability control points and using the layered and bidirectional dynamic programming algorithm,the coarse matching is processed in DSI(disparity-space image).And then the fine matching is processed in delta DSI to get the high-precision and dense disparity map.Experiments with real images show that the method can alleviate the effect of horizontal streaks caused by traditional DP-based stereo matching algorithm and improve the stereo matching speed and result.