Research Of Distributed Video Coding Based On Compressed Sensing

Compressed sensing (CS) theory can accurately reconstruct the original signal through sampling frequency which is well below Nyquist sampling frequency; thus, the complexity of the encoder is reduced. On the basis of researching the CS theory, this thesis applies CS theory in Distributed video coding (DVC).This thesis firstly researches the CS theory, and compare the performance of several reconstruction algorithms by simulating. For the shortcoming of wavelet transform in image sparse representation, contourlet transform is used as the sparse basis of image in this paper, and verify the advantages of sharp frequency localized contourlet transform (SFLCT) in image sparse representation. Then, this thesis researches the distributed compressed sensing (DCS) theory, and the joint sparse models (JSM), verify by simulating that joint reconstruction using the correlations among signals, can accurately reconstructed the original signals using fewer measurement values than conventional CS.CS and DVC can both be used to reduce the complexity of encoder, thus, this thesis researchs the DVC based on CS theory, which is named distributed compressed video sensing (DCVS). For the shortcomings of existing DCVS frameworks based on JSM-1 model in processing the intense movement video signals, a new DCVS framework based on the sparse filter correlation model (SFCM) is proposed. The simulation results show that the proposed framework provides high quality of reconstruction for intense movement and slow movement video signals. Finally, instead of wavelet transform, SFLCT is used as the sparse basis in this DCVS framework to improve the quality of reconstruction.