Face Recognition using PCA on Enhanced Image for Single Training Images

Abstract

An image enhancement based principal component analysis (PCA) method is proposed to deal with face recognition with single training image per person. The method combines the original training image is with its reconstructed image using only a few low-frequency discrete cosine transform (DCT) coefficients and then performs PCA on the enhanced training images set. In comparison with the standard eigenface algorithm and recent single training image based extended eigenface algorithms on ORL face database, the proposed method shows an improvement of more than 6% in recognition accuracy