Research On Free Viewpoint Video Coding Technique

In recent years, with the development of digital multimedia technology, the demands for interactivity of video and sensual experience are increasing. In this case, Free Viewpoint Video (FVV) comes out. Compared to traditional2D and3D video, FVV allows users to interactively control the viewpoints and can generate dynamic3D scene at any viewpoint. Moreover, FVV has remarkable advantages. It not only provides broad fields of views, stereoscopy and immersiono for users, but also make users to select views with a certain range.What’s more, it supports multiple users to watch at the same time. However, with the popularization of HD video and UHD video, as well as the necessity of depth maps for virtual viewpoint, data explosion increases. Thus, the transmission and storage of FVV system are facing new challenges. Effective compression methods are necessary. Depth image has its own features.We cannot use traditional video coding methods to encode depth map, otherwise it will introduce distortion when rendering virtual viewpoint. Therefore, the research of free viewpoint video coding becomes more and more significant.In this paper, we make research on free viewpoint video coding techniques based on the high efficiency video coding standard HEVC. In order to reduce the bit rate while achieving a better quality of the synthesized view, this paper proposes efficient techniques to compress a depth video by taking into account coding artifacts caused by depth map estimation inaccuracy and block-based coding.Besides, the computational complexity of encoder increase much due to the additional depth map. This paper proposed two kinds of fast algorithms for depth intra prediction, to save coding time while ensuring the quality of synthetic viewpoint. At last, The author of this paper has studied for a long time on the real-time encoding software named MTC265, which is based on HEVC. The author did some optimization including mode selection and quantitative for the part of intra prediction. The results show that optimized MTC265has high performance.