Evaluation of point localisation and similarity fusion methods for Gabor jet-based face verification

A comparative evaluation of two problems addressed in local Gabor feature-based face recognition is presented: localisation of points for feature extraction, and fusion of Gabor-based local similarity measures. For the former problem, three different point configurations are compared: a face-like mesh, a ( rigid) rectangular grid and a shape-driven mesh. Regarding the problem of combining local Gabor similarities for better discrimination between subjects, several state-of-the-art techniques are evaluated: support vector machines, boosting of multilayer perceptrons, sequential floating forward search, a variant of the classical linear discriminant analysis, best individual feature selection, and a closely related technique that has been recently proposed. All the experiments were carried out in configurations I and II of the XM2VTS database.