High Efficiency 3D Video Representation And Coding

3DTV has drawn much attention both in the academic circle as a new video system in recent years. While providing the audience with exciting 3D experience,3DTV also bring a big challenge to researchers in video processing and video coding field. Improving the representation efficiency and reaching higher coding performance of 3D videos are important and meaningful topics in research and development of 3DTV.This thesis first generally introduces the character of 3DTV system, the current research work done in this field and sums up in several core questions to be answered. Especially, the author investigated the standardization progress on 3D video coding and pointed out the possible prospect of new video coding standard for 3D video. Based on these investigations, several newly developed 3D high efficiency representation and coding techniques are introduced in this thesis.To improve the temporal consistency of the depth sequences in 3D video, an enhancement method is proposed. This method utilizes adaptive temporal filtering to remove the temporal noises in the depth. Experiments show that this method can highly improve the temporal quality of depth sequences and the representation efficiency of 3D content.Furthermore, this thesis introduces a new 3D video coding platform. This platform is optimized according to new demand of the development of 3D video coding techniques. Joint Rate-Distortion Optimization and View Synthesized Prediction for depth, developed by the author, are integrated into the platform. Joint Rate-Distortion Optimization uses a new R-D model to improve the R-D performance of 3D video coding and View Synthesized Prediction enhanced the inter-view prediction of depth. Experimental results show that the coding gain of this proposed 3D video coding platform is up to 9.225%.