Enriched Gabor Feature Based PCA for Face Recognition with One Training Image per Person

Gabor feature based classification approaches are widely used in face recognition, because they are insensitive to changes in illumination and facial expression. However, most of strategies only use the magnitude of the Gabor wavelet representation of images to generate feature vectors. When only single training image per person is available, the performance of these methods may be limited. In this paper, by making use of the slope angle as well as the magnitude of the Gabor wavelet response, we propose a novel Enriched Gabor feature based Principal Component Analysis (EGPCA) algorithm for face recognition with one training image per person. Experiment results show that the algorithm has better performance than other methods such as (PC)²A, E(PC)²A and SVD perturbation in a face recognition task when using the FERET database.