A novel block compressed sensing based on matrix permutation

Block compressed sensing (BCS) has great potential in image coding application for its low storage requirement and low computational complexity. However, in order to improve the reconstructed-image quality by eliminating block effect, joint reconstruction is used in the existing BCS algorithms. Thus, the computational complexity of those algorithms at decoder side is still high. To achieve a low complexity at both encoder and decoder sides while simultaneously improve the reconstructed-image quality, a novel BCS strategy with matrix permutation is introduced and investigated, in which matrix permutation is used prior to sampling to reduce the maximum block sparsity level of the 2D signal. Simulation results show that the novel matrix-permutation-based BCS approach gets a significant gain of peak signal-to-noise ratio (PSNR) of the reconstructed-images compared with the existing BCS approaches.