Abstract

Traditional direct reconstructions of CT are limited by many kinds of artifacts. To reduce image noise and artifacts, we propose an improved maximum a posteriori estimation expectation maximization (MAPEM) algorithm for image reconstruction in cone-beam CT. In this paper, first a cone-beam MAP-EM algorithm is developed for Bayesian reconstruction based upon the Markov random field in the form of Gibbs function and upon the Poisson data model. Then for solving the "one-step-late" problem of MAPEM algorithm, we deduce new reconstruction formula for cone-beam CT. The improved reconstruction method is verified by applications to simulation data and real CT data. Experiments results show this improved method is effective. Its reconstruction images are accurate and clear.

Paper Details