Multiresolution based Kernel Fisher Discriminant Model for Face Recognition

This paper presents a wavelet Kernel Fisher classifier (WKFC) for face recognition. Wavelet transform is used to derive the multiresolution based desirable facial features. Three level decomposition is used to form the pyramidal multiresolution features to cope with the variations due to illumination and facial expression changes. The Kernel principal component analysis (KPCA) method maps the input multiresolution data into an implicit feature space with a non linear mapping. The Fisher classifier is applied to multiresolution featured KPCA mapped data. The effectiveness of the WKFC algorithm is compared with different algorithms for face recognition using ORL and FERET databases. This algorithm outperforms the other existing algorithms