In region-based stereo matching algorithm, the uniform size of support-window and identical reference value will influence the matching effects. In order to solve these problems, a novel stereo matching algorithm is proposed. The proposed algorithm can adaptively select the size of support-window and distribute weights based on geometrical distance and color distance. Firstly, the support-window size was selected dynamically on the basis of image edge information. Then, the weight model was proposed, according to the change characteristics of matching values. Finally, the dense disparity map was obtained by calculating disparities point-by-point while using sum of weighted color distance as similarity measurement. Experimental results show that this algorithm is fast and efficient, and can effectively reduce the matching noise and improve the matching precision of depth discontinuities and low-textured regions.
展开▼