Reconstruction Algorithm Based On Multireference Frames Hypothesis Set Optimization For Compressed Sensing

In the traditional Nyquist sampling process,it is required that the sampling frequency of signal cannot be lower than two times of the signal bandwidth.This restriction creates lots of pressure in the encoding side,the large number of sample data also make compression procedures more difficult.All things mentioned above is unacceptable in the encoding resource constrained scene.The emergency of Compressed Sensing(CS)solved aforementioned problems.CS breaking the limits of Nyquist sampling theorem,makes the sampling and compression procedures of data in one step,which is particularly suitable in the encoding resource constrained scene.Distributed Compressive Video Sensing(DCVS)brings CS into Distributed video coding(DVC),its complexity in encoding side is less than DVC,so when DCVS was proposed,it caused much attention,then a large number of research results have been achieved.Multi-hypothesis(MH)prediction technique can obtain a high quality performance.However,the existing MH algorithm only use the adjacent Key frames as reference frames,they do not make use of the adjacent CS frames,so the reconstruction quality can be improved.Multi-Reference Frames Hypothesis Optimization(MRHO)algorithm is proposed in this paper.This algorithm expands the selection of hypothesis vectors by increasing the number of reference frames,then,the highest-correlation hypotheses,which constitutes the optimal hypothesis set,are selected.The quality of the prediction set is improved under the same size with the original hypothesis set,which means the reconstruction quality of CS frames is improved.In addition,in order to improving the reconstruction quality much further.we increased the number of reference frames for Key frames,MRHO algorithm is used to improving the quality of prediction set,and increased the quality of re-reconstruction for Key frames in the end.Simulation results show that,in low sampling rate,the reconstructed quality of Key frames and CS frames are all increased effectively by MRHO algorithm.