Improved blind image restoration scheme using recurrent filtering

In the paper, the idea of space-adaptive regularisation is introduced into the nonnegativity and support constraints recursive inverse filtering (NAS-RIF) algorithm for blind image restoration. A newly developed cost function is obtained by adding a space-adaptive regularisation term to the cost function of the NAS-RIF algorithm to prevent noise amplification and to reduce ringing artifacts. Compared to the original NAS-RIF algorithm, the improved NAS-RIF algorithm is particularly effective under low SNR conditions. Also, the experimental results show that the convergence characteristic of the improved NAS-RIF algorithm is much more stable and the converged solution is capable of providing an excellent estimate of the original image