Image restoration with an a priori estimation of the point spread function

An approach to the deconvolution of images of the Earth's underlying surface considering the distorting effect of the atmosphere is considered. The peculiarity of this approach is that the point spread function (PSF) used in the linear model of image restoration is unknown and should be estimated a priori. To this end the same image is used and information that the observable scene comprises objects with contrast brightness gradients. The Gumbel distribution of extremal values is taken as a stochastic model of image fragments with high gradients. For the alternative the family of the Johnson curves has been chosen. The Bayes decision rule based on these distributions identifies the extremal gradients. The brightness variations of a blurred image fragment along the gradients serve as a material for the PSF restoration. The image itself is restored with the use of a standard approach. An example is given to illustrate the PSF and blurred image restoration.