Research On Stereo Matching Approaches And Occlusion In Binocular Vision

Binocular vision is an important research of computer vision field. Its goal is make the computer has the ability of gaining 3D environment information from 2D images by simulating biological vision. Stereo matching research becomes a focus at home and abroad in recent years, which as the most crucial and the most difficult problem in binocular vision. The research both has important theoretical significance and application valuable. Therefore, this paper focuses on the techniques of stereo matching and occlusion detecting in binocular vision, which in order to obtain accurate and dense disparity information from the observed scene.Firstly, this paper starting with overcome the obvious stripe-like defects caused by traditional dynamic programming stereo matching algorithm, a dynamic programming stereo matching algorithm is proposed, which based on the bidirectional constraints of row and column. A new global energy cost function is constructed using the disparity discontinuity and occlusion information in the scanlines. Then a global optimization strategy based on this energy cost function is designed. The disparity discontinuity problem in column direction of scanline is solved on the basis of guaranteeing the disparity smoothness in row direction of scanline.Secondly, in order to solve the problem of the occluded pixels always affects the accuracy of the disparity image in stereo matching. A method based on the constructed energy cost function is proposed. Here, the occlusion constraint and the dynamic routing are mixed together to solve it. The information of both matched and occluded points are obtained through finding the optimal path using dynamic programming, and occlusion is detected and handled by disparity consistency testing and the occluded pixels filling, and then the dense disparity map is obtained. Finally, in order to verify the effectiveness and feasibility of the proposed dynamic programming stereo matching and occlusion detection algorithm through experiments. The disparity is calculated by use four standard stereo image pairs, and then combining with experimental results and false matching rate, the proposed algorithm and similar stereo matching algorithm are analyzed and compared.