Blind deconvolution using the maximum likelihood estimation and the iterative algorithm

Abstract The problem is considered of the blind deconvolution of an object image and a point-spread function (PSF) from a blurred image in a linear space-invariant system. A simple technique for the problem is described. This technique consists of two steps. The first step is the identification of the Fourier modulus of the PSF from the blurred image and the support information of the PSF by the maximum likelihood estimation. The second step is the deconvolution of the object image and the PSF by the iterative algorithm with the identified Fourier modulus and support constraint of the PSF. The performance of the technique is presented by computer simulations of the blind deconvolution for noncentrosymmetrical PSFs.