基于控制点约束及区域相关的立体匹配算法

立体匹配是计算机视觉的关键问题.为了得到准确匹配的稠密视差图,通过对基于特征和基于区域立体匹配算法的讨论,结合这两种算法的优点,提出一种新的基于控制点及区域相关的立体匹配算法.该方法首先在利用Harris角点检测算法检测出角点的基础上,对角点进行立体匹配得到精确的匹配点对即控制点,然后在控制点的约束下对非角点像素进行基于区域相关的立体匹配,得到整体稠密的视差图.这样既缩小了匹配搜索空间,又保证了匹配的可靠性.

Control points-constrained and area-related stereo matching algorithm

The stereo matching is a key problem of computer vision.In order to obtain dense disparity map with precise matching,a new control points-constrained and area-related stereo matching algorithm was proposed by combining the feature-based approach and area-based process method.First,corner was detected by Harris corner detecting algorithms.Then,matched the corner and obtained precise matching points that were control points.At last,matched the no-corner by area-related process method and under the constraint ...