IMAGE-RESTORATION UNDER MINIMUM OF A-PRIORI INFORMATION

A consistent systematic approach to the problem of image restoration of brurred and noisy images is presented without the use of Bayesian a priori information. Photon noise is taken into account. The unknown object is treated as a multidimensional set of parameters for which effective statistical estimates have to be obtained. The proposed approach is based on an extended notion of an admissible estimate (in the sense of information theory) and Occam's rule of choosing the simplest object which is consistent with the observations. Occam's razer is implemented by transformation to the principal components of the inverse (or most likely) estimate that are generated by Fisher's information matrix. The proposed approach can also be applied to many other inverse problems.