基于特征约束的图像点匹配算法

立体匹配是计算机视觉领域的一个关键问题,同时也是立体视觉中的难点.为了得到更好的匹配结果,给出了基于特征约束的图像点匹配算法,首先利用基于边缘特征技术得到特征点,再利用两阶段匹配算法,第一阶段主要是以特征匹配和区域匹配相结合的方法得到候选匹配点对,第二阶段对候选匹配点对进行对称性测试,以去掉部分误匹配点.实验结果表明,匹配速度较快,效果良好,有一定的使用价值.

A Matching Algorithm of Image Points Based on Characteristic-Constraint

The stereo matching is a key problem in the computer vision field,and is also the difficult problem.In order to better matching results,the image points matching algorithm is given based on the similarity of characteristic point.First,the characteristic points are obtained by using feature-based approach,then a two-stage matching algorithm is used.In the first stage,the features matching and region matching is combined to obtain matching points.In the second stage,the symmetry of selected matching points is tested and some mistaken points are removed.Experimental results show that this algorithm is effective and of certain value.