Stochastic Relaxation as a Tool for Bayesian Modeling of Astronomical Images

. Sampling techniques are used to explore the Bayesian posterior density in imaging problems. Besides the familiar MAP estimators, such techniques can easily provide more detailed information on the posterior, such as moments and marginal densities. From these, error bars and confidence levels can be assessed, and hypothesis testing performed. The algorithms are implemented in IRAF, and are being tested on both simulated and actual ROSAT images. 1. Introduction The image formation process introduces uncertainties in pixel values, due to both deterministic (Point Response Function, PRF) and random (noise) effects. The traditional approach to this problem involves some sort of inversion technique, and these are usually unable to use optimally all of the information available both a priori and in the data. An alternative approach is to adopt a data modeling perspective of the imaging problem. The work reported here aims at developing Bayesian-inference signal modeling methods based on sa..