一种基于角点引导的快速立体边缘匹配方法

提出了一种新的基于角点引导的快速立体边缘匹配方法.在匹配过程中,为了保证其快速性引入了两种新约束:角点约束与边缘约束.首先,利用匹配的角点信息引导边缘的匹配,对如何利用已匹配的角点引导边缘匹配进行了详细的阐述,同时,在引导过程中提出了只利用8个方向的区域扩张方法,进一步保证快速性.其次,提出边缘约束实现点与点的匹配,使搜索范围仅限制在几个像素内,再利用极线约束实现点的匹配,必要时可结合灰度相关约束.结果表明:该算法取得比较好的匹配效果,引入的两种约束在很大程度上提高了速度和正确率,边缘点匹配的正确率高达97%以上.

Fast stereo edge-matching algorithm based on corner-guidance

A new fast and robust edge-matching algorithm guided by matched comers was presented. In the whole matching process, the corner constraint and the edge constraint were proposed to guarantee the matching speed. First, the matched corners were used to guide the edge matching. How to use the previous matched comers to guide the edge matching was shown. Furthermore, region-growing idea was introduced to get matched edges quickly. Second, the edge constraint was proposed to limit the search area in several pixels, then epipolar constraint was used to achieved matched points, if necessary the cross correlation was also utilized. Numerous experimental results with various images clearly demonstrate that the benefit of integrating matched comers into the edge matching procedures is obvious, and with the two new constraints this method seriously improves the speed and the correct matching ratio is higher than 97%.