We demonstrate that (a) classical methods of image applications is adequate for representing images with continuous reconstruction from projections can be improved upon by considering gray levels. The potential function of the Gibbs model is defined the output of such a method as a distorted version of the original in a local neighborhood in which 3 1 3-pixel cliques as well as image and applying a Bayesian approach to estimate from it the pair cliques are used to model the continuity of borders around original image, and (b) by selecting an ''image-modeling'' prior distriimage regions and the smoothness within individual regions, rebution (one from which random samples are likely to share important
spectively. The resulting model has the property that random characteristics of the images of the application area), one can do image configurations generated (e.g., by Gibbs sampler) from better than using some other Gibbs priors. Our demonstration is from the governing distribution indeed exhibit properties shared by the area of positron emission tomography of the brain. We present images with piecewise-smooth regions. It should be noted that some encouraging results obtained using both simulated and real this model, although quite general, serves only as a demonstradata.