Gabor feature based classification using Enhance Two-direction Variation of 2DPCA discriminant analysis for face verification

This paper derives and implements a new technique called horizontal and vertical Enhance Gabor discriminant analysis (HVGD) for image representation and recognition. In this approach, we firstly use Gabor wavelets to extract local features at different frequencies and orientations from facial images. The horizontal and vertical principal component analysis (HVPCA) is then applied directly on the Gabor transformed matrices to reduce sensitivity to imprecise eye detection and face cropping. To improve upon the traditional discriminant analysis methods for face verification, the enhanced Fisher linear discriminant model (EFM) method is finally applied to further remove redundant information and form a discriminant representation more suitable for face recognition. The results show that the HVGD method performs better than the PCA, the FLD, and the EFM. The top recognition accuracy of our proposed method can reach 97.7% on the Yale database.