A segment-based stereo matching method with ground control points
Jun Xiao, Linyuan Xia, Liqun Lin, Zhentao Zhang · 2010
Most stereo matching algorithms are pixel-based. From the perspective of human stereo vision, setting segment as a processing element is more practical. A new fast stereo matching algorithm based on segmentation with ground control points (GCPs) is presented in this paper. Firstly, an object-oriented segmentation method is used to segment the reference image, which not only considers color characteristics, but also takes the spatial context of each pixel and the shape features into account. Secondly, a Winner-Take-All (WTA) matching method using SAD (Sum of Absolute Differences) algorithm has been employed to obtain the initial disparity map, and then gives a GCPs-extraction method by taking full advantage of color and gradient information. To enhance the reliability of GCPs, some heuristic methods are employed. Finally, fusing color segments and GCPs, optimal disparity images are obtained. Consistent with human being's vision, this algorithm combine segmentation and GCPs speeds up stereo matching process while keeping matching results accurate.