Bayesian and maximum entropy algorithms that use a constant balancing parameter present large residuals in the low or the high S/N regions depending on the value chosen for the parameter. We present the development of a variant of the Bayesian algorithm with entropy prior (called FMAPE) that uses a balancing parameter that is variant in space. This variation of the balancing parameter implies a variable resolution in the restoration, allowing high resolution with low bias in high S/N regions and a lower resolution with lower artifacts in poor S/N regions. We have applied the algorithm to the FOC and WF/PC cameras of the HST.
Full paper available in HTML and compressed PostScript