Research On Distributed Multiple Video Coding

The conventional video coding methods, such as MPEG and H.26x standards, use motion estimation and motion compensation to exploit statistics correlations in the source coding. But the source encoding is very complex and is not suitable for systems that need low complexity. Distributed Source Coding (DSC) is a new communication frame which has a simple encoder. It is especially useful in practices in which the encoder has low computation abilities, such as mobile camera phones, sensor network and so on. DSC with its simple encoding and robustness in transmission has received more and more attention. DSC has been received much interest from industrial and academic fields at present. Distributed Video Coding (DVC) is an application of DSC in video coding.To solve real time video communication problem in unreliable network, Multiple Description (MD) has arise more and more attention all over the world. MD is that source is divided into several descriptions by encoder firstly, then each description can reconstruct video of acceptable quality and all the descriptions together can reconstruct higher quality video. Unlike layered video coding techniques, each description generated by Multiple Description Coding (MDC) can independently be decoded and reconstructed to acceptable quality. This can give a graceful degradation of received video with loss information, while it also avoids catastrophic failure of layered coding due to loss of base layer. This thesis aims to design the schemes of efficient DVC to satisfy network communication. The research focuses on three aspects.1. Propose a novel distributed multiple description video coding scheme based on sampling, in which key frame of DVC is encoded by MD codec and transmitted over the corresponding channel. We employ three different sampling methods, time domain sampling, pixel domain quincunx and DCT domain zero padding. This scheme combines the advantage of DVC as well as robustness of MD. Experiments demonstrate that it can get better performance than some general DVC methods in switch channel. Besides, it demonstrates higher robustness in packet-loss channel than general DVC due to the MD algorithm.2. Propose a new distributed multiple description video coding scheme based on wavelet domain zero padding. The source video frames are transformed using two dimensional DWT. In DWT domain, certain numbers of zeros are padded in vertical directions and horizontal directions. Coefficients of new size are inverse transformed using IDWT and results in an over-sampled frame of bigger size. Pixels are more correlated now and then video source is divided into four descriptions by sub-sampling. The experiment demonstrates that this scheme is better than DCT domain zero padding when only one description is received.3. A DVC-based mobile communication system is constructed and a pixel DVC is realized in this system.