Stereo Matching with Global Edge Constraint and Occlusion Handling

In this paper, an effective constraint is proposed to leverage the stereo matching in early vision literature. Firstly, some particular edges are extracted to compose the new smooth constraint by categorizing the color edges into different groups. Then the optimal support windows can be established based on the proposed constraint. Finally, the disparity map would be estimated by using match propagation within the optimal support windows. Compared to the traditional window-based algorithms, our method is more effective to deal with the difficulties caused by the occluded regions or disparity discontinuities. We evaluated our algorithm with real stereo pair of images and the experimental results showed the good performance.