Stereo matching algorithm based on dynamic programming and image segmentation
Hong Xie · Applied science and technology · 2011
For the problems that the traditional stereo matching methods based on graph cuts are time consuming,and the dynamic programming stereo matching algorithm has low matching-rate and horizontal streaks of disparity map,this paper puts forward a new stereo matching method based on Dynamic Programming and Image Segmentation.First,the reference image can be divided into many regions efficiently and reliably using the method of Adaptive Multi-threshold Image Segmentation.Then,we extract the region edge,and get the parallax of the border by using multi-seed dynamic programming algorithm,and based on the unit of domain,get the parallax within the domain by using the algorithm of Graph Cuts.At last,the complete disparity map is finished by fitting the two parts.It is proven by experiments that this algorithm can improve the matching speed,and get dense disparity with high matching accuracy.