Stereo matching using iterative reliable disparity map expansion in the color-spatial-disparity space

摘要: In this paper, we propose a new stereo matching algorithm using an iterated graph cuts and mean shift filtering technique. Our estimates the disparity map progressively through following two steps. first step, with previously estimated RDM (reliable map) that consists of sparse ground control points, updated dense is constructed constrained energy minimization framework can cope occlusion. The technique employed for solution proposed model. second more accurate denser crosschecking in CSD (color-spatial-disparity) space. expands reliable disparities repeatedly above steps until it converges. Experimental results on standard data set demonstrate achieves comparable performance to state-of-the-arts, gives excellent especially areas such as discontinuous boundaries occluded regions, where conventional methods usually suffer.