Research On Natural Image Coding And Decoding Based On Compressed Sensing

Nyquist sampling theorem points out that the sampling frequency of band-limited signal must not be less than twice of the highest frequency in signals in order to reconstruct the original signal.However,with the increase of signal bandwidth and sampling frequency,it will definitely not only bring the system hardware equipment a huge challenge,but also bring broadband signal processing and transmission system enormous pressure.Compressed Sensing(CS)is a new image processing theory.It breaks through the bottleneck of Nyquist sampling theorem and makes it possible to deal with broadband signals at a low sampling rate.Since CS was raised,it quickly became an international hot spot.Natural image coding and decoding based on compressed sensing is studied in the paper.The issues studied by this paper mainly include:(1)Reconstruction algorithms of natural image based on compressed sensing are studied.Based on TVAL3,the classic TV algorithm,three improved algorithms are proposed:TV reconstruction algorithm based on Robert gradient,reconstruction algorithm based on the gradient sparse of both Horizontal Vertical and Robert,and reconstruction algorithm based on the joint sparse of gradient and wavelet.Simulation and testing show the superiority of reconstruction algorithm based on sparse of both gradient and wavelet compared with another two proposed algorithm.(2)Grey image based on block compressed sensing and its reconstruction algorithms are studied.Repeat block compressed sensing with global reconstruction program is proposed to avoid the blocking effect of block reconstruction.On this basis,non-uniform sampling is introduced and program of reconstruction based on distinct block compressed sensing is proposed for the sake of observers on the degree of concern about natural images in different regions is various.To meet the needs of the observers' subjective,this paper further puts forward the block compressed sensing of region of interest,which can greatly enhance the quality of reconstructed key region.(3)Compressed sensing applied to color images and its reconstruction algorithms are studied.Three schemes of color compressed sensing are proposed:RGB channels decoding alone,RGB channels with differential mode reconstruction and non-uniform sampling with group sparse reconstruction.The study showed that compared with RGB channels decoding alone,RGB channels with differential mode reconstruction have a higher reconstruction quality at the same sampling rate.The group sparse,using wavelet domain as the sparse representation domain,is applied to the color compressed sensing in order to better use of the correlation between RGB three channels.Non-uniform sampling scheme is adopted by distributing different sampling rate in the large coefficients of wavelet and small coefficients of wavelet to further improve the quality of the color image reconstruction.The research work shows that the global reconstruction algorithms based on multiple sparse domain can high probability to reconstruct the original image under a low sampling rate,and the introduction of non-uniform sampling for block compressed sensing can effectively improve the quality of reconstructed key region,and the system of color images based on block compressed sensing which combined with non-uniform sampling idea and reconstruction algorithm of group sparse has good performance.