BAYESIAN IMAGE-RESTORATION - AN APPLICATION TO EDGE-PRESERVING SURFACE RECOVERY

Bayesian methods are used to recover a 2-D surface by assuming that th ere is a textural image that can be modeled by a Markov random field a nd that the original surface is composed of different surfaces, each o f which is associated with one textural state. Both parametric and non parametric methods are used to enforce smoothness of these surfaces. I terative procedures are examined for simultaneous restoration of the t extural image and estimation of underlying parameters. From the estima ted textural image and the estimated parameters, an estimate for the o riginal surface can be obtained. Some numerical examples are presented , including an illustration from electron microscopy.