Automatic face recognition via wavelets and mathematical morphology

Presents a new method for automatic face recognition and verification. The proposed approach is based on a two stage process. In the first step a wavelet decomposition technique or morphological nonlinear filtering is used to enhance intrinsic features of a face, reduce the influence of rotation in depth, changes in facial expression, glasses and lighting conditions. Preprocessed images contain all the essential information for the discrimination between different faces and are next a subject for learning by a modified high order neural network which has rapid learning convergence, very good generalization properties and a small number of adjustable weights. The system is not based on task dependent geometric feature extraction, and as such, it can be easily applied to other image recognition tasks