Research On Fast Multi-View Stereo Video Coding Algorithm Based On 3D-HEVC

As the video is an important means of information dissemination by humans, the video codec technology is developing forward rapidly. In order to meet the demands of authentic and interactive video, the video technology is to develop in the direction of high-definition, multiple viewpoints. But with the improvement of video resolution, as well as the increasing number of viewpoints, it takes serious challenges to the network bandwidth and processor capacity. Therefore, in order to solve the problem, the international telecommunication union(ITU-T) and international electrotechnical commission(ISO/IEC) form a Joint Collaborative Team on Video Coding(JCT-VC) in 2010 for development of the high efficiency video coding(HEVC) standard, which is released in early 2013. Meanwhile, a Joint Collaborative Team on 3D Video(JCT-3V) is formed to study the next generation of 3D video coding standard, which will be released in the future. According to the experimental results, it can achieve about 50% time saving compared with the previous generation standard, i.e., the multi-view video coding(MVC) standard. However, this standard is difficult to be applied, because the computational complexity of it is very high. Therefore, while maintaining the coding performance, how to effectively reduce the encoding computational complexity is got the high attention of the academia and industry both at home and abroad.Firstly, this paper describes the key technology of HEVC briefly, and presents the comparative analysis with H.264/AVC. Then, the key technology of 3D-HEVC and the research status at home and abroad are introduced in detail. Also, the selected topic significance is presented.Aiming at the high computational complexity of the flexible dividing block technology in 3D-HEVC, a fast parameter selection algorithm based on the encoded information is proposed. It utilizes the spatiotemporal correlation, inter-view correlation, and inter-component correlation to design a variety of prediction strategy, which can effectively reduce the splitting depth levels of coding units and transform units, and the prediction modes. The experimental results show that this algorithm can reduce about 68% encoding time compared with the original encoder.For the difficult problem about the computational complexity of 3D-HEVC, the parameter selection problem of the encoder is modeled as a classification problem. According to choosing the appropriate feature vector, an accurate Bayesian classifier is designed. Through the parameter selection process of the online learning based Bayesian classifier, it can effectively classify the encoding parameter set. The experimental results show that this algorithm can achieve about 55% time saving with no quality loss. Therefore, these two algorithms proposed in this paper have a certain reference value to spread this standard in the international multimedia industry.