Architecture solution of the non blind iterative Lucy- Richardson deconvolution algorithm

Abstract

This paper addresses the problem of blind image restoration. The non blind iterative Lucy-Richardson (LR) algorithm is considered. The proposed architecture solution consists of making the LR algorithm totally blind and fully automatic. For this aim, first a proposed point spread function estimation method is used to estimate the blur kernel and make the LR algorithm blind. Moreover in order to make the LR algorithm automatic, the iterative deconvolution process is controlled by a proposed stopping criterion based on a no reference blurred image quality metric that suggested in [1]. Computer simulations using the Gblur LIVE database [2] are provided in order to illustrate the proposed architecture effectiveness. A comparative study with the shock filter blind deblurring method [3] is performed, emphasizing the good proposed architecture solution behavior.