Bayesian object identification

This paper addresses the task of locating and identifying an unknown number of objects of different types in an image. Baddeley & Van Lieshout (1993) advocate marked point processes as object priors, whereas Grenander & Mille r (1994) use deformable template models. In this paper elements of both app roaches are combined to handle scenes containing variable numbers of object s of different types, using reversible jump Markov chain Monte Carlo method s for inference (Green, 1995). The naive application of these methods here leads to slow mixing and we adapt the model and algorithm in tandem in prop osing three strategies to deal with this. The first two expand the model sp ace by introducing an additional 'unknown' object type and the idea of a va riable resolution template. The third strategy, utilising the first two, au gments the algorithm with-classes of updates which provide intuitive transi tions between realisations containing different numbers of cells by splitti ng or merging nearby objects.