Distributed Compressive Video Sensing

Distributed video coding combined with compressive sensing is a new technology. Comparingto the exsiting disttibuted compressive video sensing system, the proposed system can improve thereconstructed video quality and the stability.Firstly, the existing algorithm of distributing measurement rate to image blocks in a CS frame isstudied. The existing algorithm is impvode in the article. At the sending end, a CS frame is dividedinto serval image blocks, whose width and height is equal. This adaptive measurement rateallocation for CS frames is implemented by making the absolute values of DCT efficients of eachimage block in a CS frame fall in different ranges. These blocks will be divided into three typesbased on the number of their measurements. At the transmitting end, different encoding methodswill be selected based on the depending block types. At the receiving side, depending decodingmethods will be selected for reconstruction of the measurements. Comparing to the performance ofexisting algorithms, simulation results show that the proposed algorithm reduces the proportion ofextremely sparse block and improve the reconstructed video quality.Secondly, the algorithm to calculate the motion intense parameters of corresponding blocksbetween adjacent frames is put forward in the paper. The image blocks are divided into differentmotion types based on the parameters. Based on the reconstruction quality from to high at thereceiving end, the scenario is resepectively defined as scenario I、scenario II. Furtherly, theadaptively divided GOP algotirhm in different scenarios is proposed in the paper, depending on thechanges of motion types of the adjacent frames. At the transmit side, the frames are divided intoGOPs with the proposed algorithm. Compared with the performance of existing fixed GOPalgorithm, the simulation results show that the proposed algorithm improves the reconstructed videoquality and the stability of the reconstruction.Finally, this paper presents a TV reconstruction algorithm combines with the side informationframe generated by the forward reconstructed key frame and the backward reconstructed key frame.TV reconstruction algorithms require a higher initial choices, so the algorithm CS frame sideinformation frame as the initial values. The method of caluating similarities between thereconstructed frame and the side information is proposed. The similarities between thereconstructed frame and the side information and those between the forward reconstructed frameand the backforward reconstructed frame are respectively set as the iteration stopping criteria for the proposed algorithm. So the TV algorithm combined with side information frame is proposed.Compared with TVAL3algorithm and the traditional GPSR algorithm, the simulation results of theproposed algorithm show that the proposed algorithm could improve the reconstructed video quality,reduce the reconstruction time and improve the speed of reconstruction.