Restoration of optical imagery with the expectation-maximization algorithm

Abstract

The expectation-maximization algorithm, as formulated in 1977 by Dempster, Laird, and Rubin [1], is a numerical procedure for computing constrained maximum-likelihood estimates of unknown parameters that influence measured data. It’s use for image restoration from quantum-limited data was first suggested by Shepp and Vardi [2] in 1982 in the context of determining concentrations of radioactive substances within the human body; applications and extensions of this method for estimating radionuclide distributions have grown enormously in the intervening years, with contributions by numerous investigators.