Depth Map Optimization Coding For 3D Video

With the continuous development of mobile communication technology and display technology, 3D video has been widely concerned and researched by academia and industry since it could able to provide a more realistic visual experience for the audience. At present, 3D video coding contains the texture map and the depth map which has a large amount of data, and video compression efficiency requirements are very high. In order to solve these problems, the 3D video coding standards are formulated. In the depth-based 3D video system, a virtual video of any viewpoint can be rendered by utilizing the virtual view synthesis. As a result, the quality of depth map coding will directly determines the quality of synthetic view. Depth map reflects the distance relationship between the object and the camera, and the drastic changes on the edge of the object would be shown in the 2D map as a small sharp edge is surrounded by flat area or slowly changing large area. This article mainly around the depth map coding to carry out the following research.⑴Coding optimization based on the depth filtering technology: Impacting on the quality of synthesis, the sharp boundary decides coding efficiency directly. After filtering processing, high frequency information is abandoned, which led to the sharp boundary becoming blurred, thus affect the depth map coding and quality of synthetic view. In order to solve this problem, the optimized filtering technology is proposed.⑵ Coding optimization based on the depth no synthesis error model(DNOSE): In the process of the 3D video displaying, depth map make observer obtain the effect of stereo video by synthesis different viewpoints texture map. In the view synthetic process, each pixel in a refer view is projected into the virtual view and falls onto a floating point coordinate. In most of view synthesis algorithms, this position is rounded to the neighbor position of an integer sample raster. So several closed pixels can be mapped to the same pixel position on the synthesis process. According to this thought, the DNOSE model is established, and the relationship of pixels in the spatial-temporal domain is determined by spatial distance, texture similarity and time distance. A depth spatial-temporal filter is proposed based on spatial-texture-time, and the strength of the filter is optimized processing.⑶Coding optimization based on the quality enhancement of the depth view synthesis: For synthesis of visual quality, the sharp boundary affects very big, and flat area affect is small, which has not need too much bitrates in flat regions. Aiming at this phenomenon, the depth map sampling coding algorithm is proposed. In order to make the sample more uniform over the surface of the object, an interlaced down-sampling approach is applied among different view. In order to better solve the problem of less reference information in the up-sampling, the reference information of across viewpoint is introduced. In the case of the same bitrates, the proposed method is better than other methods for quality of rendered view.According to the simulation experiment for the above algorithm, the depth map coding efficiency and synthetic visual quality are improved, and realizes the 3D video coding depth map optimization.