Face recognition using Euler Principal Component Analysis

Face images with visual variations can significantly influence the performance of a face recognition system. Euler Principal Component Analysis (e-PCA) uses a dissimilarity measure to increase the differences between subjects even though the face images are under the influence of visual variation. Previous experiments show that e-PCA is particularly effective in reconstructing occluded face images. Thus, in this paper, we investigate if e-PCA can be used to solve the problem of visual variation in face recognition by using the reconstructed face images for the classification process. Different classifiers are also used in our investigation to examine the effect of the reconstructed face image data on the process. Experiments are done on ORL, AR and Yale face databases and it shows that there are improvements in the recognition rate using e-PCA under certain circumstances.