一种利用动态规划和左右一致性的立体匹配算法

立体匹配是计算机视觉领域研究的一个重要课题,为了得到准确、稠密的视差图,提出了一种利用动态规划和左右一致性的立体匹配算法。该算法首先分别以左、右图像为基元图像,计算各自的视差空间图像,在视差空间图像上利用动态规划,计算得到左视差图和右视差图;然后通过使用左右视差图之间的一致性关系,消除误匹配点,得到较为准确的部分视差图;最后利用视差图的顺序约束关系,给出未匹配视差点的搜索空间计算方法,并利用一种简单有效的方法来计算这些点的视差值。在一些标准立体图像对上所做的实验结果表明,该算法效果良好。

A Stereo Matching Algorithm Using Dynamic Programming and Left-right Consistency

Stereo matching is one of the most important researches in computer vision.In order to obtain dense and correct disparity,a stereo matching algorithm using dynamic programming and left-right consistency is presented.Firstly,the left and right disparity space images are computed using the left and right images as basic image separately.Secondly,in the computed disparity space images,the disparity images are computed using dynamic programming.Then the left-right consistency of the disparity images is used to remove the mismatching pixels,and to generate the part of near real disparity images.At last,according to the ordering constraint of disparity image,a method detecting the searching space of unmatched pixels is presented.And an ordinary but efficient strategy is proposed to finalize these pixels.The experiments on some standard stereo pairs are executed,and the results show the algorithm is effective.