Research On Coding, View Warping And Motion Compensation Design For Three Dimensional Television

Three Dimensional Television (3DTV) has the ability of making audience feel right on the scene. 3DTV is the future evolution. The advance of Autostereoscopic displays and signal processing algorithms make the solid foundation. Research on 3DTV is of substantially theoretical, practical and commercial meanings.A practical 3DTV system generally utilize video plus depth information. Notably 3DTV has to deal with multiview video streams. This paper makes research on the field of 3DTV coding, rendering method and motion compensation hardware implementation. The achievements are as follows:(1)Proposed a Multivew Video plus Depth Coding method, which adopts Hirerarchical B structure and view warping on the temporal and spatial domain respectively. The method differentiates views into main and auxiliary group, and transmits disocclusion information to help decoder reconstruct quality views. The method also provides the ability to couple main view's depth image period with auxiliary view's group of picture length to exploit spatial correlation.(2)Proposed a depth-aided inpainting algorithm for disocclusion restoration during rendering. The method preprocesses the depth image with dilation. The restoration order is based on the block priority. The metching cost function utilizes texture and depth similarity to choose the most metching block. Compared with other methods, proposed method has superiority in geometic fidelity, structure propagation and depth differentiation. It has the ability to provide efficient pictures for predictive coding.(3)Proposed a parallel serial mixed luma and 3 stage recursive chroma interpolation architecture. Through serial processing the vertical input data, the design optimized interpolator data flow. Also proposed hardware adaptive algorithm for refrence picture management, which accelerates the motion compensation process with small cost.