Blind Deconvolution Using a Monotonicity Constraint on the PSF

Abstract

It is well recognized that blind deconvolution is a severely ill-posed problem and proper constraints on the image and the system point spread function (PSF) should be applied to counteract the ill-posedness. In this paper we investigate a novel PSF constraint, the monotonicity, which means the value of the PSF monotonically decreases (or does not increase) from the center of its support. We regard the monotonicity as a common property of many simplified but well accepted PSF models, such as the geometrical model of defocus, the Gaussian model and the synthetic model in astronomical imaging. The property is utilized as a PSF constraint in a novel iterative blind deconvolution algorithm RL-CLSE. Experiments on real microscopic data show that the proposed constraint can significantly improve the quality and stability of blind deconvolution.