Bayesian image processing of data from constrained source distributions. I: Non-valued, uncorrelated and correlated constraints (Q1820543)

A series of Bayesian image processing algorithms which incorporate various classes of a priori source information in treating data which obeys Poisson and Gaussian statistics is derived using maximum entropy considerations. The standard maximum likelihood equations are shown to be a special case of Bayesian image processing when the a priori information about a source distribution \(\{\phi_ j\}\) is solely that a non- vanishing probability for each element value \(\phi_ j\) exists only in some finite interval, \(a_ j\leq \phi_ j\leq b_ j.\) Bayesian image processing equations of the a priori source information that all \(\phi_ j\) are finite \(-\infty <\phi_ j<\infty\) and each \(\phi_ j\) distribution has a defined mean \({\bar \phi}_ j\) and a defined variance \(\sigma_ j\) are derived. The Bayesian image processing equations are also derived when the a priori source information is that all \(\phi_ j\geq 0\) and that each \(\phi_ j\) distribution has a defined mean \({\bar \phi}_ j\) and a defined variance \(\sigma_ j\). The a priori source distribution constraint that a correlation exists among nearby elements is also considered. The results indicate improvement over standard methods.