Research Of Distributed Video Codec Technique Based On Compressed Sensing

The conventional signal processing theorem states that the video signal can be reconstructedwithout distortion only if the sampling frequency is greater than or equal to two times the highestsignal frequency.This sampling algorithm always involve high computational complexity.Recently,a new sampling algorithm, called compressive sensing (CS), is proposed which breaks thebandwidth limitation of the Nyquist sampling theory. Hence, CS has broad market prospects inimage or video processing. In the CS, if the signal has sparsity prior, it can be compressed using anunderdetermined matrix that is unrelated with the transform matrix.Then, with a high probability,the compressed data can be reconstructed by solving a minimization problem. CS cansimultaneously achieve video sampling and compression.In this paper, we focus on implementing the CS algorithm into the distributed video encoding.Firstly, we establish a measuring rate based reconstruction model when studying the state of artdistributed video encoding technology. The proposed reconstruction model pays more attention tothe following problem: the different inter-frame correlation. The reconstruction model can find therelationship between the measuring rate of frames and the reconstruction quality of videos.Secondly, we propose an adaptive measuring rate allocation algorithm.The paper has proved theeffectiveness of the proposed reconstruction model and the measuring rate allocation algorithmthrough experiments. Thirdly, we proposed an improved Motion-Compensated Block-basedCompressed-Sensing Smoothed-Projected-Landweber (MC-BCS-SPL) video reconstructionalgorithm through studying the CS based video reconstruction method. The proposed videoreconstruction algorithm can further exploit the inter-frame correlation in video through improvingthe intial frames. This proposed video reconstruction algorithm can obviously enhance the PSNR ofthe reconstructed video sequence.