A Novel Stereo Matching Method based on Rank Transformation
Wenming Zhang, Kai Hao, Qiang Zhang, Haibin Li · 2013
For the problems that the traditional local SAD algorithm is affected by the central pixels largely and noise easily, this paper puts forward a new Rank-transform stereovision matching algorithm based on window weighted function. Firstly, the traditional nonparametric Rank transformation is used for left and right gray images to suppress the noise; meanwhile, the weighted window function is added to Rank transformation to keep the detail information of the scene as integral as possible; the match based on SAD algorithm is applied in gray space and the second match on the exponent constraint of color difference is applied in color space to remove mistake matching points; finally a dense disparity map can be obtained. Experimental results show that the proposed algorithm can suppress the noise much better than the traditional SAD algorithm. At the same time, the accuracy rate is nearly improved by 5.6%.