基于小波及动态规划的相位匹配

提出了一种新的基于小波变换及动态规划的相位匹配算法。该算法采用金字塔式的多尺度匹配结构，顶层和中间层利用小波相位信息建立扫描线对的视差空间图，并引入动脉规划的方法，在获得全局最优化的匹配路径的同时有效地解决了遮挡问题。为提高匹配结果的置信度，该算法结合特征匹配确定了一系列高置信度的匹配点对作为匹配路径的控制点。金字塔底层则直接利用小波相位差和瞬时频率更新由上层由插得到的视差图，保持了相位匹配的亚像素精度。立体像对测试结果表明：该算法改进了现有的相位匹配算法；在匹配时间和区配精度上优于包含动态规划的区域匹配算法。

Phase Matching Based on Wavelet Transform and Dynamic Programming

A new phase matching approach was proposed based on wavelet transform and dynamic programming. This approach constructs a multi scale pyramid matching structure based on wavelet transform. On the top layer and each intermediate layer, dynamic programming technique is introduced into the disparity space image of each scan line to obtain the global optimized matching path with proper occlusion detection. To improve the accuracy of the matching results, feature matching following the high confidence first principle is used to determine GCPs (ground control points) to force the matching path to make large disparity jumps that might otherwise have been avoided because of large occlusion costs. Phase matching runs directly on the bottom layer to retain sub pixel accuracy of the disparity. Compared with the current phase matching approaches and block matching approaches based on dynamic programming, the experiments demonstrate much better performance of the proposed algorithm.