Occamian approach in the image restoration and other inverse problems

摘要

AbstractA non‐Bayesian approach to inverse problems is discussed in terms of the image restoration problem. The approach is based on the extended notion of the feasible estimate and on the Occam's principle of choosing the simplest object, consistent with the data. The Occamian estimation is performed by transforming the inverse or maximum likelihood estimate to its principal components, which are induced by Fisher's information matri