A Study of Local Features for Unconstrained Parts Based Face Classification

Recently it has been shown that face verification based on a unconstrained parts approach, comprised of local feature extraction and a mixture of Gaussians classifier, is considerably more robust than holistic approaches in terms of translations (which can be caused by automatic face localization) and pose variations. In this report we first investigate whether features based on Local Principal Component Analysis (LPCA) are more discriminative than the often used features based on the 2D Discrete Cosine Transform (2D DCT). We also investigate several methods for modifying the two feature extraction techniques in order to counteract the effects of linear and non-linear illumination changes, without losing discriminative information. Experiments on the XM2VTS database show that the performances of 2D DCT and LPCA techniques are similar, suggesting that the 2D DCT technique is preferable due to its lower computational complexity. Furthermore, contrary to previously published results, the experiments show that the use of deltas (differences in DCT/LPCA coefficients obtained from neighbouring blocks) of low-order coefficients can adversely affect robustness