Initialization and Optimation of Deformable Models

The deformable model literature has in general been very focused on the formulation and development of new models or the solution of a specific application. Teh final and crucial steps of initialization and optimazation of the deformable model, needed for making inference, have received very little attention.

We present an improved general method for initialization based on a filter interpretation of the likelihood. This method will, for a large number of models and problems, generate the initialization significantly faster than the previously proposed general method. To perform the optimization a number of general algorithms from the optimization literature is introduced, which are novel in the context of deformable models.

These algorithms are compared to general algorithms previously used for optimization of deformable models. Tests are performed on real data.