Refinement using O Scene Reconstruction

An occurrence of occlusion is the challenge problem in stereo matching. In this paper, we propose a refinement algorithm of disparity maps with occlusion handling. In order to refine the disparity map, we obtain an initial disparity map with an optimization algorithm based on the modified constant-space belief propagation low complexity. An occlusion map is computed using unique constraint from initial left and right disparity map. We classify occlusion into two types from obtained occlusio suitable occlusion handling process, respectively. Proposed handling process based on potential energy function extends the disparity values of the visible pixels to the occluded pixels. Experimental results show that the proposed algorithm confirmed to be valid and can be widely used for 3D scene reconstruction.