Comparison of Image Restoration Techniques
A. D. Storrs, R. L. White (STScI), P. C. Thomas (Cornell)
A full interpretation of undersampled, high signal to noise images is possible if the system response (point spread function, or PSF) is known. Iterative deconvolution techniques are in regular use in the astronomical community, for example in determining the color-magnitude diagram for dense cluster cores. The usual deconvolution programs (clean, maximum entropy, and Lucy) are generally optimized to resolve a field into point sources.
Planetary objects are rarely point sources. The items of interest in an image generally consist of smooth brightness variations on the field of a disk, which may or may not have a sharp limb. We present the results of using several techniques on the same image. We use one of the Hubble Space Telescope (HST) images of asteroid 4 Vesta (Zellner et al. 1996, BAAS v. 28, p. 1100). This image combines both an irregular, sharp limb and some spatial extent and albedo variation, with a system who's PSF is very well known. We compare the results of the sigmaCLEAN, Wiener, Lucy, and MEM packages in IRAF/STSDAS, and a new damped Lucy method run under IDL.
Although these methods can be used to enhance the visibility of
features that can be seen in the raw images, it is not entirely
clear if the spatial resolution of the marginally sampled raw
image can actually be increased. The observer is left to compare
the results and judge for themselves.