Illumination invariant face recognition

Abstract

We present a novel method for face recognition by enhancing the quality of the input face images, which may be too dark due to different lighting conditions. We propose to extract the FFT features or the dual-tree complex wavelet (DTCWT)-FFT features from the enhanced face images and use the Support Vector Machine as a classifier. Our experiments show that our proposed methods compare favourably to the FFT features without image enhancement, and the methods in [1] and [10] for the Extended Yale Face Database B and the CMU-PIE face database. © 2013 Springer-Verlag.