Data acquired with the charge coupled device camera on the HST are modeled as an additive Poisson-Gaussian mixture, with the Poisson component representing cumulative counts of object-dependent photoelectrons, object-independent photoelectrons, bias electrons and thermoelectrons, and the Gaussian component representing read-out noise. Two methods are examined for compensating for read-out noise. One method is based upon approximating the Gaussian read-out noise by a Poisson noise and then using the expectation-maximization (modified Richardson-Lucy) algorithm for Poisson distributed data to effect the compensation. This method has been used for restoring HST images. The second method directly uses the expectation-maximization algorithm derived for the Poisson-Gaussian mixture data. This requires the determination of the conditional-mean estimate of the Poisson component of the mixture, which is accomplished by the evaluation of a nonlinear function of the data. The second method requires more computation than the first, but modest improvements in the quality of the restorations are realized, particularly for fainter objects.
Full paper available in HTML and compressed PostScript