High–level Image Priors in Confocal Microscopy Applications

The incorporation of high–level prior information in Bayesian imaging has been proposed via the theoretical frameworks both of Grenander and Miller (1994) and Baddeley and Van Lieshout (1993). Recent developments in Markov chain Monte Carlo algorithms for varying dimension problems (Green, 1995) have made the application of these high–level models more widely accessible, although this remains more complicated than using the usual low–level Markov random fields models. One application area in which incorporating information on the object level in a scene would be of particular benefit is in confocal microscopy. In this paper we outline two such applications, the first addresses uncertainty in segmenting confocal microscope images, and the second is an object recognition problem.