Research On The Viewpoint Synthesis Optimization And Error Concealment Of 3D-HEVC

The next generation high efficiency video encoding standard HEVC adds a lot of new technologies on the basis of retaining the structure of traditional hybrid encoding framework and it can save almost half coding bit with the same constructed video quality. Three-dimensional video(3DV) can give the audience a more intuitive and vivid expression, so 3DV encoding technology is currently concerned because of its strong market. The moving picture expert group(MPEG) developed a new generation 3D-HEVC of 3DV encoding standard by HEVC as the benchmark, and the multi view video plus depth(MVD) was used as the data structure for 3DV.Depth information is used to synthesize the virtual viewpoint in the decoder. Depth encoding uses view synthesis optimization to calculate distortion in the cost function and it need to synthesis virtual viewpoint between known viewpoint. Virtual view synthesis technology is depth-image based rendering(DIBR). When warp to the virtual viewpoint from the known viewpoint, there will be some error mapping points in the case of the occlusion or exposure. In this paper, we explore different weights to error mapping points in the calculation of the distortion, making the view synthesis optimization is more reasonable, so that the depth mode selection can chose the better one. Special processing for those points can make the higher quality of virtual viewpoint in the decoding end also can improve the efficiency of encoding depth.The error concealment for the multi view plus depth encoding structure is very limited. In this paper, we study the selection of candidates based on the consistency criterion which consider the consistency of time and the consistency between different viewpoints. The criterion is based on the similarity of the matching block between the adjacent time frame and the adjacent viewpoint frame. When the viewpoint of the coherence is calculated, the consistency between different viewpoints of occlusion point in mapping process is not calculated. The effect of the occlusion point is ignored, so that the consistency of the viewpoint is more accurate. According to the intensity of motion, the degree of similarity between the time and the viewpoint is weighted. The improved the error concealment algorithm can get better video quality and give users better feeling.