基于双序列比对算法的立体图像匹配方法*

在分析现有立体匹配方法的基础上，提出一种基于双序列比对算法的立体图像匹配方法。将立体图像对中同名极线上的像素灰度值看做是一对字符序列，使用基于动态规划思想的双序列比对算法对这些对字符序列进行匹配，以获取立体图像视差。为验证该方法的可行性和适用性，采用人脸立体图像对进行实验。实验结果表明，使用该方法进行立体图像匹配能获得光滑的、稠密的视差图。基于动态规划思想的双序列比对算法能够有效地解决立体图像匹配问题，从而为图像的立体匹配提供了一个实用有效的方法。

Stereo image matching approach based on pair-wise sequence alignment algorithm

This paper presented an approach for stereo image matching based on pair-wise sequence alignment algorithm after analyzing the existing stereo image matching methods. Pixel gray value on homonymy epipolar in stereo image pairs was thought of as a pair of character sequences, and then the disparity of stereo image pairs could be acquired by the pair-wise sequence alignment algorithm based on dynamic programming to matching these pairs of character sequences. To demonstrate the feasibility and applicability of the proposed approach, adopted face stereo image pairs to the experiment. Experimental results show that using the proposed approach for stereo image matching can produce a smooth and dense disparity map. The pair-wise sequence alignment algorithm based on dynamic programming can effectively solve the stereo image matching problem, so it provides a practical and effective approach to stereo matching of image.