目前,立体匹配是计算机视觉领域中最活跃的研究主题之一。为了快速并更精确的对特征点进行立体匹配,本文提出了一种新的基于特征点的立体匹配算法。该方法独立于特征点的检测算法,先以扫描线作为匹配单元,然后以鲁棒函数为匹配代价函数,最后用顺序约束对每一匹配单元的视差图进行检测与校正。实验证明,该方法的匹配精度高于传统的基于NCC(norm alized cross-correlation)的立体匹配算法,并且运行时间快,可以应用于纯软件的基于特征点的立体视觉系统中。

Stereo matching is currently one of the most active research topics in domain of computer vision.A new feature-point-based stereo matching algorithm is proposed in this paper,in order to match feature points more efficiently and accurately.It is independent of feature detection algorithms,using the scan line as the matching unit,using the robust function as the matching cost function and using the ordering constraint to detect and correct errors in each matching unit in the disparity map.Experiments show that the matching accuracy of this algorithm is better than the traditional NCC(normalized cross-correlation) based stereo matching algorithm and the running time is quite fast.So,it can be used in the independent hardware feature-point-based stereo vision system.