Study On Distributed Video Coding

Distributed Video Coding (DVC) is a new paradigm for video compressing, based on Slepian-Wolf and Wyner-Ziv multi-sources coding theorems.Compraed with traditional video coding methods, the DVC-based architectures present the following functional benefits:flexible allocation of the overall video codec complexity, improved error resilience, codec independent scalability and exploitation of multiview correlation. These benefits make the DVC has important application value in many wireless video applications like wireless low-power surveillance, wirelsee multimedia senses network and so on.This dissertation focuses on studying and analyzing several key modules of distributed video coding, and the main research and contributions are listed as follows.1.The distributed source coding algorithms have been studied, which is the foundation of the DVC.A distributed coding algorithm based on the q-ary low-density parity-check accumulate (Q-LDPCA) codes was presented to solve the problem of coding two correlated non-binary sources with asymmetrical statistical dependency. On one hand the use of Q-LDPCA codes in the proposed system makes coding without converting non-binary sources to binary bits feasible, and thus it overcomes the reduction in correlation caused by conversion.On the other hand it makes coding rate varying adaptively with correlation of the correlated sources with asymmetrical statistical dependency. Experiment shows that the proposed algorithm performs better than coding by binary LDPC code with conversion first.2.A joint bit-planes decoding (JBPD) algorithm for distributed video coding is proposed. The JBPD algorithm uses the decoded bit-plane results of quantization coefficients as prior knowledge, and decodes the other bit-planes to avoid correlation weakening between sources when decoding bit-planes independently. It makes full use of correlation between Wyner-Ziv frames and key frames;therefore the rate-distortion performance of distributed video coding can be enhanced without increasing decoding complexity. Simulations indicate that, compared with the independent bit-planes decoding system, the proposed algorithm can achieve up to 16.5% bit-rate saving under the same condition, and performance improves more apparently with the increase of quantization levels.3.The modeling of virtual correlation channel and frame reconstruction of the DVC have been studied. An adaptive frame reconstruction algorithm for distributed video coding is proposed. The new algorithm reconstructs the quantization samples of Wyner-Ziv frames adaptively according to the parameter of the correlation model,quantization interval and corresponding values of the side information frames based on the correlation model between Wyner-Ziv frames and side information frames.It can improve the subjective and objective quality of decoded Wyner-Ziv frames compared with the conventional fixed boundary reconstruction method. It is also more flexible than improved boundary reconstruction method with correcting coefficient got by experience. Simulations indicate that, at the same bit rate up to 0.91dB coding gain of PSNR can be obtained with the proposed algorithm to the fixed boundary method when the other modules are all the same in the system.4.The algorithms of side infromation frame construction have been studied. Several widely used algorithms are implemented and compared, and a side infromation frame construction algorithm based on varying block size is proposed.