Research On Decoder Motion Estimation Algorithms In Distributed Video Coding

As the development of mobile communications, wireless networks and multimedia technology, the range of video application is becoming wider and wider, many video systems require the terminal can realize not only low complexity decoding, but also real-time encoding and transmission, such as mobile visual telephone, wireless camera, wireless low-power surveillance system. With the restriction of equipment cost, cell capacity and so on, the terminal's ability of calculation and storage is limited, which makes it can't do complicated video compression.Distributed video coding(DVC) is an efficient path to solve the problem of resource limitation and it's the application of DSC in video coding.It adopts intra-frame coding and inter-frame coding, which shifts the computational burden from the encoder to the decoder.As the shift doesn't sacrifice compression performance theoretically, it satisfies the demand of video coding and communication.DVC is based on channel coding, which has the feature of simple encoding, low transmission errors, high compression efficiency and better robustness.it offers a new thread to transmit video signal robustly.DVC is a new field in the research of video compression, one of the primitive and innovative technology of DVC is the generation of side information. The accuracy of the evaluation of the side information determines the RD performance of DVC, and affects its development.Firstly, this paper presents basic principles of DVC. Secondly, it analyses the framework of DVC, then describes the importance of side information in DVC.Thirdly, focuses on the theory of side information generation, and puts forward an improved scheme based on unsupervised learning of motion vector which improves system's performance.The scheme generates and optimizes side information through the introduction of rate control at the encoder and EM model at the decoder. Lastly, a DVC system is implemented based on the scheme, its principle of work is presented. A MATLAB simulation system has been developed for the DVC system. By comparing with typical side information generation methods, the outstanding performance of the proposed side information generation scheme is justified, and the validity and advantages and feasibility of the proposed DVC system are demonstrated.