Study Of The Key Technologies For Distributed Video Coding

Distributed Video Coding (DVC) is a new coding paradigm based on two major informationtheory results: the Slepian-Wolf (1973) and Wyner-Ziv theories (1976). It adopts intra-frame codingand inter-frame coding, which shifts the computational burden from the encoder to the decoder.DVC is mainly used to solve major issues, such as high encoding complexity, poor scalability andfault-tolerant, which existed in video communication by traditional video coding. The developmentof DVC opens up a new challenging opportunity to improve the traditional video codec. Thisdissertation introduces the basic theories and typical coding architecture in DVC, and focuses onstudying and analyzing side information generation and rate control in DVC.First, considering the motion intensity inequality at different regions, it is proposed thatclassifies image blocks into two modes: fast motion and slow motion by simply computing theDiscrete Cosine Transform (DCT) coefficients at the encoder, and the decoder chooses differentalgorithm to generate side information according to the corresponding motion mode. Simulationresults demonstrate the proposed scheme, without increasing the encoder complexity can not onlyachieve efficient compression, but significantly improve the subjective and objective quality of sideinformation.Secondly, considering the rate control algorithm using a feedback channel would increase thedecoding complexity and cause decoder-to-encoder transmission delay, it is proposed an improvedrate control algorithm. The proposed algorithm adopts the block classification principle presented inthis paper, and the mode value is transmitted to the rate control module in order to simplify the rateallocation, and this paper uses the adaptive Laplace factor in correlation noise model (CNM) tocontrol rate in realtime according to the size of motion vector. Simulation results demonstrate theproposed rate control algorithm obtains better RD performance at the cost of slightly increasedencoding complexity, and avoids using the feedback channel, which can guarantee real-time videocommunication.Finally, based on the Wyner-Ziv video coding system proposed by Stanford University, animproved Wyner-Ziv video coding system without feedback is constructed, which adopts the aboveside information generation and rate control algorithms. The experimental results show thatcompared with the existing DVC without feedback, the proposed scheme not only achieves efficientcompression, but significantly improves the subjective and objective quality of the reconstruction image. Furthermore, the encoding speed is considerably improved compared with H.263inter-coding. The proposed DVC system greatly decreases the encoder complexity and can berealized easily.