Joint-MAP reconstruction/segmentation for transmission tomography using mixture-models as priors

A Bayesian method, including a pointwise prior comprising mixtures of gamma distributions, is applied to the problem of transmission tomography. A joint MAP (maximum a posteriori) procedure is proposed wherein the reconstruction itself, as well as all pointwise parameters, are calculated simultaneously. It uses an algorithm that successively refines the estimate of the mixture parameters and the reconstruction. The approach aims to solve the problem of low counts statistics in transmission tomography. Initial simulation results with anecdotal reconstructions show that the gamma mixture model likely outperforms the ML (maximum likelihood) method and FBP (filtered-backprojection) algorithm.