一种基于仿射不变区域约束的立体匹配算法

针对现有立体匹配算法匹配速度的不足，提出一种基于区域约束的快速立体匹配算法。选取4对稳定的特征点作为初始匹配点对，在左右图像中分别构建四边形，然后将四边形划分成4个三角形，以每个三角形对为待匹配区域，利用仿射不变性进行区域内特征点匹配。获得特征点匹配对后，再以每个特征点为中心进行开窗，利用传统的基于灰度信息的方法进行精度验证。实验表明，该算法在保证一定精度的情况下能获得较高的匹配速度，可满足实时性要求。

Stereo matching algorithm based on affine invariant region constraint

Aiming at the exiting stereo matching algorithms which could not meet the speed requirements, a novel quick stereo matching method based on region constraint was presented. Firstly, four pairs of stable feature points were selected as the initial matching points, and each quadrilateral was structured according to the four pairs of matching points in the left and right image respectively. Then the quadrilateral was divided into four triangles and each triangle was an area for matching. Feature points in the area were matched based on affine invariant. After the matching feature points were obtained, a small window was structured centered on those points. At last, the precision testing was implemented using the traditional method based on gray information. Experimental results demonstrate that the proposed algorithm can notably improve the matching speed which meets the requirement of real time.