A General Scheme for Training and Optimization of the Grenander Deformable Template Model

General deformable models have reduc edthe need for hand crafting new models for every new problem. But still most of the general models rely on manual inter action by an expert, when applied to a new problem, e.g., for selecting parameters and initialization. In this pap er we prop ose a full and uni ed scheme for applying the general deformable template model proposed by Grenander et al. [7, 13] to a new problem with minimal manual inter action, beside supplying a training set, which can be done by a non-expert user. The main contributions compared to previous work are a supervise d learning scheme for the model parameters, a very fast general initialization algorithm and an adaptive likeliho odmodel based on local means. The model parameters are traine dby a combination of a 2D shap elearning algorithm and a Maximum Likeliho odbased criteria. The fast initialization algorithm is based on a search appr oach using a lter interpr etation of the likelihood model.