Generic methods in pattern recognition and their application to image processing

Recently, a study has been carried out in the framework of a project, funded by the Fund for Scientific Research (FWO), related to new types of pattern classifiers, based on Bayesian decision functions. The study aimed at optimising the time performances of the classification algorithms (if necessary at the cost of a increase in the time, necessary to estimate the unknown model parameters from the training set). The study yielded a model, based on the approximation of the class conditioned probability densities, using an expansion in a special non-orthonormal basis. Although the usefulness of the model was demonstrated at the end of the FWO project, we were able to derive from this study a number of new and interesting research subjects. This project is meant to include the mentioned research subjects, without excluding supplementary subjects with would emerge from the execution of the project.