Face Recognition Under Varying Illumination Based on Ordinal Features

Illumination variation is one of the most difficult problems for face recognition. In this paper, we represent a new ordinal feature based method for face recognition under varying illumination. We employ 2-D wavelet transform to compress face images and extract ordinal features from them. Here, the ordinal feature is extracted more easily than before and competent for face recognition, which is invariant for illumination changes. Experiments are carried out on Yale B face database, and experimental results demonstrate that our proposed method achieves satisfying recognition rate under illumination variations.