A Weighted l0 Shearlet-based Method for image deblurring

This paper proposes a weighted l(0) shearlet-based model for image deblurring. The main purpose of this work is to further exploiting the sparsity of the recovered signal. In order to achieve this goal, a generalized gradient regularizer is introduced in the weighted shearlets based model. The added regularizer suppresses artifacts effectively. The spit Bregman algorithm is employed to update the multiscale-based weighted matrix in every iteration. By updating the weighted matrix, the solution information in present step can be transmitted to next iteration, an iterative-feedback algorithm framework is produced. Experimental results show that this method yields significantly improvement in PSNR. However, the numerical experiments also show that more computing time is required due to the utilization of the redundant shearlets system.