Reseach On Compressive Sensing-based Video Compression And Coding Technology

In the field of video compression coding, most of those schemes consist ofcomplex encoder and simple decoder, the processing of spatial and temporalredundancies for data compression causes that the computational complexity atencoder is five to ten times more complex than it at decoder, which makes thoseschemes unsuitable for real-time processing of high-speed video. Recently proposedcompressive sensing theory offers the possibility to solve the conflict betweenperformance and quality. According to the theory, for signals with sparsity, thesampling rate required to achieve high-quality reconstruction could be much lowerthan the lower-bound provided by Shannon sampling theorem. Video signal is sparsesignal, compressive sensing theory could be used for processing.This paper mainly focuses on analyzing and researching the application ofcompressive sensing theory in the field of video compression coding. On the basis ofthe introduction of compressive sensing theory and the applications of it as well astraditional video codec technology, various video coding schemes based oncompressive sensing are discussed and compared with each other deeply. Further, areal-time coding scheme for high-speed video is proposed. That is using relativelycomplex off-line decoding process to exchange the simplicity of online encodingprocess. At encoder, in order to satisfy the need of high-speed real-time encoding, nocomplex pre-processing is performed, and fast online sampling encoding is directlyprocessed on video data. At decoder, three-dimensional complex dual-tree waveletused as transform domain, and make an improved AMP algorithm the reconstructionalgorithm to recover data. Experiment shows that, this scheme fully take advantage ofthe intra-redundancy and inter-frame correlation of video. By ensuring the recoveryaccuracy of high-speed video while effectively improving the online coding efficiency,it is proved suitable for high-speed video.