Research On Disparity Control And Virtual View Rendering For Autostereoscopic Display

Autostereoscopic video has a widely application. There is a greater challenge to autostereoscopic video about acquisition, transmission, display etc. Many researchers have developed many video processing algorithms to adapt different screen in the autostereoscopic video processing, reconstruction and display. However, there is less research about disparity control of convergence autostereoscopic video. Meanwhile, the virtual viewpoint rendering still exists many problems, such as hole, occlusion, distortion and boundary error etc. This paper carries out detail research works aiming to disparity control of autostereoscopic video and virtual viewpoint rendering.In the aspect of disparity control, this paper studies camera parameters and autostereoscopic video disparity control technology. According to the basic principle of binocular disparity and space mapping, we obtain the corresponding points of zero disparity at different viewpoints, and thus computing the disparity between the different viewpoints. By using this disparity values, viewpoints are translated, and the disparity control of convergence autostereoscopic video is achieved. Root mean square error(RMSE) and subjective evaluation model(MOS) are used to evaluate the performance of convergence autostereoscopic video after disparity control. Experimental results show that the proposed disparity control method can improve the performance of convergence autostereoscopic videoIn the aspect of virtual viewpoint rendering, this paper studies the depth map preprocess and boundary refinement technology. The different equipment and scene cause misalignment what appears in color map and depth map boundary. The expansion and corrosion are used to detection misalignment, and the Canny operator is applied to detect boundary. By adjusting the boundary, the misalignment problem is restrained. At last, VSRS 1D Fast software is used to synthesis virtual views. Experimental results show that the proposed method is validity to improve the quality of the virtual view rendering.