Algorithm Research On Hypothesis Set Design Based Distributed Compressed Sensing

In 1924,Nyquist deduced that the highest bit rate in an ideal low pass channel. In order to recover the analog signal without distortion, the sampling frequency should be no less than 2 times the maximum frequency of the analog signal spectrum, which is also known as the sampling theorem. However, in the case of some resources- limited encoder(such as energy, storage devices, etc.), sampling at the rate of two times the highest frequency of the spectrum is difficult to achieve and sustained. Compressed sensing(CS) which integrates the sampling and the comp ressing stage of traditional sampling is a new sampling technique, breaking the limits of N yquist sampling theorem. The high efficiency compression and very simple operation make it particularly suitable for processing multimedia video data in the encoding resource constrained scene. Then, CS is applied to the field of video coding and decoding by compressed video sensing(CVS). Both in terms of computational complexity and data amount, the burden on the encoding side are greatly reduced.In spite that a lot of research results have been achieved on the CS based video scheme, the reconstruction performance at low sampling rate is still not satisfied. Excellent reconstruction performance can be obtained with high sampling rate, but this is a great burden on the memory of the encoding storage device and the calculation is overhead. This is contrary to the original intention of using CS theory, and it is also very difficult to achieve in the bottom of the wireless network with some energy limited nodes.In this paper, we improve multi- hypothesis prediction based distributed compressed video sensing(DCVS) codec by proposing hypothesis set update(HSU) algorithm and dynamic reference select(DRS) algorithm. The quality of the reconstructed video is improved, without increasing the complexity of the encoding. HSU algorithm, using the method of constructing the dictionary, updates the original hypothesis set. The algorithm replaces hypotheses with low correlation to the target block in original hypothesis set by the hypotheses of high correlation searched in the extended hypothesis set. These operations make the target block obtain better sparse representation, and make use of the spatial correlation of the frame, and then improve the quality of reconstruction.On the basis of the two reconstruction, the DRS algorithm selects more suitable reference frames for the current frame which makes full use of inter- frame correlation and improve the reconstruction quality. The algorithm improves the reconstruction quality of non-key frames by exploiting distributed compressed video sensing and HSU and enhances that of key frames by DRS at the decoder, thus achieving a higher decoding qualit y at low sampling rate, besides the method performs well when the video content changes fast. Simulation results show that the proposed block based distribute compressed video sensing codec combined HSU and DRS algorithms(HD-BDC VS) system can effectively improve the quality of video decoding at low sampling whether the content of video alters quickly or not.