Research Of Distributed Compressive Video Sensing

Distributed compressive video sensing is a novel video coding scheme which combines compressive sensing and distributed video coding. This scheme becomes a research hotspot attracting many scholars at home and abroad because it reduces the energy consumption of the video capture terminals and improves the compression efficiency. The main work of this paper is to improve the quality of side information, the quality and the efficiency of signal reconstruction.First of all, to cope with the situation of inconsistent movement in same image block during side information generation, this paper proposes an optimal generation method which utilizes weighted motion compensation to deal with overlapped blocks. The experiments prove the proposed method improves the side information quality with a better subjective vision.Second, to improve the accuracy of atom selection rules in subspace pursuit during signal recovery, this paper proposes a modified subspace pursuit with residual side information. This algorithm selects atoms by their influence on temporal interdependency of frames instead of their energy. It is proved via experiment that the proposed algorithm outperforms the original one and keeps the efficiency unaffected at a low sampling rate.At last, to fulfill the real-time requirement of video information, this paper proposes a fast recovery algorithm based on sparsity estimation to improve the recovery efficiency. This algorithm estimates the sparsity of the residual between key frame and non-key frame by normalized hard threshold decision for residual side information. The simulation shows, compared with the original methods, the proposed method significantly improves the reconstruction efficiency, and also has some improvement in reconstruction quality at a low sampling rate.