Research Of Video Codec Technique Based On Distributed Compressed Sensing

Today, video application in the network has been becoming increasingly widespread, and the requirements of its transmission speed and clarity have become more sophisticated with the rapid development of communication technology. But at the same time, the occupied consumption of the physical storage space and the usage of transmission bandwidth also have been increasing. The H.26x and MPEG technology are two kinds of mature video processing framework which need the higher sampling rate and the more complicated transformation, quantization and the entropy coding techniques, and all of these are causing a great burden to the encoders. Distributed technique can finish the work of independent encoding and joint decoding, so we can transfer the complexity from encoders to the decoders, but it also causes a great of waste of sampling points, the compressed sensing technique is an adaptive solution to the above problems.The compressed sensing (CS) technology can build a unique sparse model according to the sparsity of the signal itself and recover it with the appropriate observation matrix, so that we can achieve the accurate reconstruction even with a low sampling rate, and its advantage becomes particularly evident especially when dealing with the one-dimensional or the two-dimensional signal. However, the real-time property of the CS technology is weak, and the delay is obvious in codec processing of the video signal. Here we introduce the distributed compressed sensing technology, which can deal with joint reconstruction at the decoder, so that we can improve the recovery efficiency of the signal significantly. Therefore, in order to solve the problem of high sampling rate to the terminal equipment and the bandwidth usage in network transmission, we combine the distributed compressed sensing algorithm with an appropriate video processing system model, which is also the main research direction in this paper.We mainly describe the basic theoretical framework of compressed sensing and introduce several sparse group, observation matrix and the corresponding recovery algorithms. On the basis of the traditional distributed video codec system combined with the CS technology, we introduces three joint sparse model of the DCS theory mainly about the JSM-1, based on which we do the research of the video codec problem, and achieve precise reconstruction of the signal at a low sampling rate consumption and less storage space occupied as far as possible. For further improvements, We present a new adaptive video processing framework with the theory of side information under the evaluation criteria of PSNR, with which we can dynamically adjust the selection of key frame in the sequence of video images to ensure the real-time characteristics are not affected and get a better recovery quality.