Block Compressed Sensing Image Reconstruction Based on SL0 Algorithm

By applying smoothed l₀ norm (SL0) algorithm, a block compressive sensing (BCS) algorithm called BCS-SL0 is proposed, which deploys SL0 and smoothing filter for image reconstruction. Furthermore, BCS-ReSL0 algorithm is developed to use regularized SL0 (ReSL0) in a reconstruction process to deal with noisy situations. The study shows that the proposed BCS-SL0 takes less execution time than the classical BCS with smoothed projected Landweber (BCS-SPL) algorithm in low measurement ratio, while achieving comparable reconstruction quality, and improving the blocking artifacts especially. The experiment results also verify that the reconstruction performance of BCS-ReSL0 is better than that of the BCS-SPL in terms of noise tolerance at low measurement ratio.