A quality measure based stopping criterion for iterative deblurring algorithms

Blind image deblurring algorithms (BIDA) constitute a subset of image restoration algorithms used to solve ill-posed inverse problems, which can be challenging in low SNR situations. Those iterative methods usually impose some regularization upon the deconvolution process in order to constrain the problem and reduce the size of solution space. In this paper, we propose to use a new stopping criterion for iterative deblurring algorithms based on our previously published no reference blur image quality measure. The rationale behind the proposed stopping criterion is to control the deconvolution process by estimating the reminding blur quantity. For test, two iterative deblurring algorithms are considered. The Lucy Richardson [1,2] and the Shock Filters methods [3,4]. The proposed adaptive approach has been tested on blurred images from LIVE database (Gblur).