The most popular and effective methods for the restoration of HST data are non-linear and hence redistribute flux in a way which often degrades the photometric content of the data. They also cause artifacts in the background to appear when large numbers of iterations are used, and the number of iterations which will give best results is often difficult to estimate. Several studies (e.g., Cohen 1988, Linde &Spännare 1993) have suggested that a workable compromise for point source photometry is to use restored images, which show better contrast and object discrimination, to identify objects and then to perform photometry by using standard PSF fitting packages (e.g., DAOPHOT, Stetson 1987) on the unrestored data. Other work (Busko 1993) has shown that there is often a tradeoff between photometric accuracy in the measurement of point sources and the quality of the background in the resultant restored image.
Many astronomical images do not contain a uniform spread of spatial
frequencies. Instead they consist of many point sources and a
smoother background intensity
distribution. Standard restoration methods do not use this
information and hence cannot significantly sharpen point sources without
causing the background to break up into speckles. Other methods, such as
those based on Maximum Entropy considerations, can impose smoothness but
generally underestimate point source fluxes.
We have adopted a different approach which separates the known point sources,
which are forced to be -functions, from a background which is forced
to be smooth. Hence we obtain the best of both worlds by
using extra information
about the image. As a further generalization, the method may be used to
iteratively find the PSF using a form of ``blind iterative
deconvolution.'' A description of the logical steps in the development
of this method and more detailed mathematics
are given elsewhere in this volume (Lucy 1994). In this paper
several illustrations of its use on different classes of problem are
presented along with details of the implementation. The test images are
taken from those prepared by the STScI Image Restoration Project.
This technique suggests several new possible observational projects using both space and ground based telescopes.