Fast Disparity Estimation Based On Multiview Video Coding

Nowadays, the computer sciences, electronics, communications technology is developing rapidly. As one of the most active fields of computer research, the multimedia encode technology is developing with the request of the application. With the movie reying of"Avatar", people are attracted by the real effect of the 3D scene, which brings the development of many other related industries, such as 3D movies, 3DTV, 3D animation and so on. Multiview video coding( MVC ) as the basis for these industries is got more and more attention.Because of the limited network bandwidth, video coding technology has been in the pursuit of low bitrate and high coding efficiency. Multiview video is multi-angle video sequences of a scene which are shot from multiple angles simultaneously. We can see that multiview video is a very large amount of data, therefore, how effectively compressing multiview video would appear to vital. To improve the compression efficiency of MVC, many algorithms have been proposed. Among all the schemes for MVC, the one proposed by HHI has achieved the best performance over the simulcast anchors. As we know that the computational complexity of motion estimation for video coding is already very high, the MVC scheme using hierarchical B pictures further increases the computational complexity due to the additional disparity estimation( DE ). To achieve real-time encoding, the computational complexity should be reduced. This paper focus on the fast disparity estimation of MVC, as well we applied the proposed new technology to the Direct mode of MVC.3D Video as the expansion of 2D video can provide the depth of visual experience to users. Now, 2D texture + depth representation has been applied in 3DTV and FTV, we can see that the depth maps as the supplementary information are permitted to be transmitted to the decoder for supporting the virtual viewpoint and 3D scene rendering in the standardization of MVC. The idea of this paper is making use of the 3D warping technology with the help of depth maps to redefine the starting point of disparity estimation, then reduce the search range, which achieve the objective of reducing the computational complexity. Experimental results verify that the proposed algorithm can greatly save the time for DE with negligible loss of coding efficiency.In this paper, the 3D warping technology is further extended to the application of Direct mode in MVC. In direct mode, the motion information which is derived from the adjacent macroblock can not be transferred to the decoder. We replace the disparity vector predicted by adjacent macroblock with the disparity vector predicted by 3D warping technology. We hope this new Direct mode can improve estimation accuracy, then improving the coding efficiency. Experimental results show that this new method can make a slight increase based on maintaining the original coding efficiency.