Research On Multiview Video Coding

With the rapid development of3D technology, video displaying technology has been developed from2D to3D, and multiview video has attracted considerable attention re-cently. Multiview video consists of video sequences captured by an array of cameras, which can be from different angles and locations, therefore, multiview video can make scene more vivid and accurate. However, the data of multiview video is very large, it’s very difficult to storage and transport, so it’s a key issue for compressing the multiview video.In this dissertation, we focus on the techniques for multiview video coding, mainly studied multi-hypothesis distributed multiview video coding, compatible efficient stereo video coding and robust multiple-description multiview video coding. This thesis gains the following research achievements:1) We propose a multi-hypothesis distributed multiview video coding. This scheme can reduce the complexity of encoder, avoid the communication between different cam-eras and utilize side information efficiently. In addition, Bayesian multi-hypothesis joint conditional probability density function is presented, which is also used in the LDPCA decoder and optimal reconstruction for the proposed scheme. The experimental results show that the proposed scheme can improve the coding efficiency.2) Compatible stereo video coding is proposed. Stereo video is special case of mul-tiview video, which includes left view video and right view video, and is used widely. However, current encoders can not compress multiview video conveniently, which are mainly for2D video. Therefore, compatible stereo video coding schemes, based on flex-ible prediction mode and adaptive prediction mode respectively, are proposed in this the-sis. The efficiency of the proposed scheme is given by the experimental results.3) Multiple description multiview video coding is proposed. In traditional multi-view video coding, once there is an error of transmission or reconstruction, the error accumulation will be caused. Multiple description coding, which is robust, can resolve this problem. First, for the multiview intra coding, multiple description multiview intra coding with randomly offset quantizers and uniformly offset quantizers are presented. Then, for the multiview inter coding, the multiple description multiview video coding is proposed, which is based on interpolating compensation preprocessing. This method can reconstruct missed block and erred block more efficiently and with better quality. The robust and efficiency of the proposed scheme are demonstrated by experimental results.