A generalized form of the Lucy-Richardson restoration method is described in which images in the object plane are divided into two channels. One of these contains point sources (i.e., delta-functions) and the other models a smooth background distribution. The latter image is regularized by the use of an entropy term and hence has enforced smoothness. This approach avoids problems encountered in the photometry of the results of non-linear restoration methods. Examples of the use of this technique on standard simulated HST data frames are given, including some assessments of its photometric accuracy. Implementations running under both IRAF and MIDAS are available.
Full paper available in HTML and compressed PostScript