Conclusions

Use of the Poisson approximation of the function that accounts for read-out noise does not introduce significant errors in the restoration of HST images. However, some improvement in photometric performance can be obtained at the expense of computation time by a more accurate evaluation of this function. The improvement is seen most for faint objects when 50 iterations of the expectation-maximization algorithm are used to perform the restorations. For star cluster fields, performing more iterations of the expectation-maximization algorithm yields significantly improved restorations, and the Poisson approximation is not improved noticeably by a more accurate approximation of the function.