Research On Technologies Of Distributed Multi-view Video Coding

With the development of multi-view video technology,more and more multi-view video applications are springing up. For some emerging applications such as smartphone video communication, multi-angle video surveillance system, traditional multi-view video coding technology has been unable to meet their requirements. Distributed multi-view video coding is just proposed to provide an effective solution for the above mentioned applications. The advantages of low complexity encoding and no communication of views greatly reduce the cost consumption of the terminal acquisition devices, and distributed coding theory has proved that applying the way of independently encoding and jointly decoding can also receive the same compression efficiency as traditional video coding does. In recent years, distributed multi-view video coding is gradually becoming an active focus in multi-view video coding research areas. Therefore, research on distributed multi-view video coding are of both theoretical significance and practical value.This thesis through the investigation, analysis and improvement of existing distributed multi-view video coding frameworks, aims to design a series of novel distributed multi-view video coding frameworks, not only to achieve compression performance improvements but also to ensure low encoding complexity. The main work and innovations of this thesis are:1. Propose an improved distributed multi-view video coding framework. The enhancement is performed on the existing DMVC framework. In order to meet the applications requirements of 3DTV, we introduce the depth sequences into our coding structure. At the encoder, we apply a block classification with low computing complexity to effectively exploit the statistical correlation between timeline frames. At the decoder, we employ 3D warping and motion vector projection techniques to further exploit the correlations of multi-view video sequences, and then propose a novel high quality side information generation and fusion method. Simulation results show that the proposed advanced multi-view distributed video coding has received both improvements of quality of generated side information and the rate distortion performance, compared with an existing framework.2. Propose a novel distributed multi-view video coding framework based on compressive sensing. Since compressive sensing and distributed coding have a natural commonality, we use compressive sensing technology to implement distributed multi-view video coding. At the transmitter, in order to make full use of sampling resources and save sampling rates, a block-based adaptive sampling rate allocation method is proposed. In each block, we sample video signal according to allocated sampling resource, and then transmit measurements to receiver for signal recovery. At the receiver, we first produce redundant dictionary and side information, taking full advantage of correlations between multi-view video sequences, and then, solve a side information constraint optimization problem to recover the original frames. Simulation results show that each proposed technique of this framework has afforded a significant contribution to the improvements of recovery performance.