FULLY BAYESIAN-ESTIMATION OF GIBBS HYPERPARAMETERS FOR EMISSION COMPUTED-TOMOGRAPHY DATA

In recent years, many investigators have proposed Gibbs prior models t o regularize images reconstructed from emission computed tomography da ta, Unfortunately, hyperparameters used to specify Gibbs priors can gr eatly influence the degree of regularity imposed by such priors and, a s a result, numerous procedures have been proposed to estimate hyperpa rameter values from observed image data, Many of these procedures atte mpt to maximize the joint posterior distribution on the image scene. T o implement these methods, approximations to the joint posterior densi ties are required, because the dependence of the Gibbs partition funct ion on the hyperparameter values is unknown, In this paper, we use rec ent results in Markov chain Monte Carlo (MCMC) sampling to estimate th e relative values of Gibbs partition functions and using these values, sample from joint posterior distributions on image scenes, This allow s for a fully Bayesian procedure which does not fix the hyperparameter s at some estimated or specified value, but enables uncertainty about these values to be propagated through to the estimated intensities, We utilize realizations from the posterior distribution for determining credible regions for the intensity of the emission source. We consider two different Markov random field (MRF) models-the power model and a line-site model. As applications we estimate the posterior distributio n of source intensities from computer simulated data as well as data c ollected from a physical single photon emission computed tomography (S PECT) phantom.