Multiview Video Coding Technology Research

With the recent development of information technology, computer graphics, computer vision, and multimedia, 3D visualization and interactivity has become available. Consequently, Multi-view Video Coding (MVC) becomes a research hotspot in the video coding. These multi-view videos are acquired by using a set of cameras synchronously capturing the same scene but from different viewpoints. As the number of cameras increases, the amount of data and redundancy rapidly increase, too. In order to store and transmit these multi-view videos for the practical applications, efficient multi-view video coding (MVC) is highly desirable. To this end, the Joint Video Team (JVT) has developed a reference model, the Joint Multi-view Video Coding (JMVC), based on H.264/AVC video coding standard. Although the coding performance of JMVC is good enough, but at the cost of very high computational complexity. Therefore, how to reduce the computational complexity while maintaining almost the same video coding quality is a key issue for any practical MVC codec.In order to reduce the computational complexity and improve the temporal and spatial random access performance of hierarchical B pictures coding structure, which has been adopted as the basic coding structure in Joint Multi-view Video Coding (JMVC), an improved prediction structure is proposed in this paper. The frame in B-view which uses the immediate previous frame as its reference adopts only temporal prediction, and inter-view prediction is skipped in non-anchor frames in P-view, so that computational complexity is reduced and random access ability is improved. The position of I-view is also studied to minimize the loss of coding efficiency introduced. Experimental results have shown that compared to the hierarchical B pictures coding structure in JMVC, the proposed scheme achieves a remarkable reduction of encoding time on average with insignificant gain loss. Meanwhile, better random access performance has also been obtained.To further reduce the computational complexity, the proposed improved prediction structure is cooperated with mode decision optimization in this dissertation. Experimental results have shown that compared with the reference prediction structure and exhaustive mode decision in JMVC, the adopted algorithm achieves a great reduction of computational complexity while maintaining almost the same video coding quality.