Wavelet Based Illumination Invariant Preprocessing in Face Recognition

Performance of face recognition system has not been satisfied due to the change of illumination on facial image. Thus, there were many proposals that dealing with illumination compensation in face recognition in the past decades. In this paper, we propose a wavelet based illumination invariant algorithm based on the illumination-reflectance model. The proposed method aims to remove illumination component. This can be done by setting the coefficients in wavelet approximation subband to zero values. The algorithms are tested on YaleB database and CMUPIE database, with encouraging results.