PSF ESTIMATION FOR IMAGE DEBLURRING

This paper addresses the problem of estimating the point spread function (PSF) of a spatially invariant linear system through which an image has been blurred. This is done by searching for specific features of the original image, such as sharp edge elements or microtexture, from which the blur effect can be identified and the PSF estimated. The original image is represented by a Markov random field which models these features. We develop a hierarchical algorithm using a multiresolution methodology. Some examples of estimation are included to illustrate the discussed method. © 1991.