Estimating a global shape model for objects with badly defined boundaries [mammography application]

Determining the shape of image features that are the result of grouping processes is a difficult task. This paper proposes a global shape model based on one or more closed polygons. The model is discussed in the framework of the general concept of pattern theory which gives important methodological guidelines for the construction of the model as well as its solution. This framework makes it possible to apply the Bayesian theory to highly structured and arbitrarily complex model classes. The determination of the optimal configuration is by maximizing the posterior probability in configuration space. Apart from the solution a measure of its reliability is also given