Maximum entropy methods and dark field microscopy images

Abstract

Maximum entropy methods are examined for the removal of noise from dark field electron microscope images. A number of maximum entropy algorithms, including χ 2 and E 2 constraints, are examined. Failure of conventional maximum entropy methods led to the inclusion of an autocovariance constraint in the algorithm. Under this constraint the noise removed from the images has a more realistic spatial distribution than under conventional constraints. This constraint, when combined with a simulated annealing algorithm, led to successful removal of noise from small test data sets.