Research On Distributed Video Coding Based On Bayes Compressive Sensing

Against traditional video encoders for high complexity can not be applied to computational and resource constrained applications(such as wireless multimedia sensor networks).Distributed Video Coding(DVC)technology has entered the relevant research Line of sight.The distributed video coding framework adopts the scheme of key frame independent coding and non-key frame reconstruction,which realizes the purpose of computing complexity from coding to decoding.It is very suitable for the field of video transmission with limited coding resources.In the distributed compression-aware video codec system,the sparse representation and the edge information directly affect the reconstruction quality of the video sequence.In order to improve the accuracy of video reconstruction,this paper improves the edge information synthesis method.In order to further improve the quality of reconstruction,for the video frame in the coding side of the sparse representation of the relevant improvements.The research contents of this paper are as follows:(1)In order to improve the quality of edge information and the quality of video sequence reconstruction,a classification weighted edge information generation algorithm is proposed based on the traditional block-based edge information generation scheme.Firstly,the blocks are classified by the correlation difference of the different blocks in the two adjacent key frames before and after the reconstruction.Then,the forward and backward motion estimation and backward motion estimation are generated for the adjacent key frames.Candidate edge information,and then set the weighting factor of the candidate edge information according to the classification decision result,and finally generate the side information,and then completes the reconstruction of the non-key frame.This algorithm makes full use of the difference of the correlation between different sub-blocks of the video frame,improves the quality of the edge information,and improves the reconstruction precision of the non-critical frame.(2)In order to obtain a better sparse description of video frames,a distributed video compression sensing framework based on clustering sparse representation is proposed by combining block sparse representation and non-local similarity between video frame blocks and blocks.Firstly,the image block is found in the current frame by the search algorithm.Then,the image blocks are composed of similar image blocks,and the sparse representation of the set is obtained by using the adaptive redundancy dictionary to obtain more accurate sparse representation coefficients.Finally,the image block is reconstructed and the reconstructed result is decomposed into the corresponding position of the image block.All the block reconstruction results constitute the reconstruction result of the video frame.Experimental results show that the proposed method can improve the frame frame sparseness and video frame reconstruction accuracy.