The quality of the point spread function is the primary factor limiting the quality of image restoration for the HST. This has been demonstrated using several types of PSF for HST WFPC image restoration: (1) an empirical PSF extracted from the image to be restored, (2) an empirical PSF retrieved from the WFPC PSF library (Baggett and MacKenty 1992), and (3) a model PSF computed using the Tiny TIM software (Krist 1993). Similar results were obtained for HST WFPC images by Krist and Hasan (1993) who attribute the inaccuracies of the Tiny TIM models to incomplete knowledge of the HST primary mirror zonal error map.
Unfortunately, none of these methods yields an ideal PSF. The empirical PSFs contain noise and non-zero background levels. Frequently it is difficult to find a good PSF star in the image to be restored, and a PSF image that is well-enough exposed to show the details in the halo will often have a saturated core. Spoiler stars must be removed from the PSF image, a procedure which can itself introduce artifacts into the PSF. In using the WFPC PSF Library, care must be taken to find well-exposed images. Some images were taken with very short exposures where the effects of finite (and non-linear) shutter travel speed cannot be neglected. In addition, use of Library PSFs can introduce artifacts into the restoration owing to the known time variability of focus. Model PSFs do not reproduce all features of observed PSFs accurately, and short-term focus variability is not currently taken into account.