Distributed Video Coding Research Based On Compressed Sensing In WMSN

Traditional video encoding and decoding technology(H.264x/MPEG-4)requires complex motion estimation and motion compensation operations at the encoding,which directly leads to high complicated calculation at the encoding.Therefore,it is not suitable for application scenarios with resource shortage at the encoding,such as wireless multimedia sensor network(WMSN)devices.Distributed compression video sensing(DCVS)is the most effective technique to solve the problem.The technology is the compressed sensing(CS)theory and distributed video coding(DVC)technology with the combination of a novel video code scheme,this scheme not only full filled the compressed and sampling at the same time,also the complicated calculation of motion estimation and motion compensation from the code to decode,control the coding of the computation.Therefore,this video encoding scheme is very suitable for WMSN devices with limited resources at the encoding.This paper through the analysis of the present DCVS reconstruction algorithms in video sequence facing challenges on the basis of quality,to improve the DCVS system for sports more violent video sequence to reconstruct the rate-distortion performance of poor problem as the research target,on the structure characteristics of video sequence between adjacent frames and CS observation prediction residual error characteristics are studied,and then focuses on DCVS reconstruction algorithm and video compression perception of quantitative method in the system of two key technologies,the main work and research results were as follows:(1)For the reconstruction algorithm in the current DCVS system,the reconstruction quality of the slow-moving video sequence is good,while that of the violent video sequence is not ideal.To solve the problem,the paper improves the reconstruction algorithm of video compressed sensing on the basis of image group sparse representation reconstruction algorithm(GSR).Firstly,a method is proposed to construct similar block groups by using the structural similarity of adjacent frame signals and the non-local correlation of inter-frame signals.Then,in order to get more matched similar blocks,this paper combined the structural similarity(EGSSIM)based on extended gradient operator with Euclidean distance as the matching criterion to generate similar block groups.In addition,during the simulation experiment,it was also found that the reconstruction quality of video sequence would be improved with the iteration of reconstruction algorithm,and the image texture,details and other information would be enriched gradually.Therefore,in order to prevent error updating the current information,this paper proposes an optimal adjustment criterion for the number of similar blocks and applies it in the iterative process of reconstruction algorithm.Simulation results show that the proposed frame based on extended gradient structure similarity between group sparse representation reconstruction algorithm(E-GSSIM-InterF-GSR)is the current mainstream video compression performance in the reconstruction perception reconstruction algorithm is improved,the specific performance in the reconstruction of the peak signal-to-noise ratio(PSNR)and based on extended gradient operator of structural similarity(E-GSSIM)than SSIM-InterF-GSR、MRMH、MH-BCS-SPL algorithm has the obvious improve,especially for football is more violent,the more obvious of improvement.PSNR has a maximum of 3.02 dB improvement,and E-GSSIM has a maximum of 0.1035 improvement.(2)In this paper,an improved DPCM non-uniform quantization method was proposed to reduce quantization distortion on the basis of the prediction residual characteristics of observed values.The improved quantization scheme improves the compression rate of the observed data on the premise of guaranteeing the reconstruction quality.In addition,based on the improved inter-frame DPCM-NSQ method,an adaptive optimal quantization deep model(AOQDM)is designed in this paper.Simulation results show that the proposed quantization scheme improves the rate-distortion performance ofDCVS system without increasing the amount of computation at the coding.To be specific,the rate distortion performance curve of the quantized scheme in this paper almost coincides with the optimal rate distortion performance curve(Best-NSQ),and even exceeds the optimal rate distortion performance curve.