An efficient method for face recognition under varying illumination

Principal component analysis (PCA) is a classical method which is often used for human face representation or recognition. However, for those images under uneven lighting conditions, the performance of PCA degrades greatly. In this paper, an efficient method for human face recognition under varying illumination is proposed. In our method, a local normalization technique is applied on the image point by point, which can efficiently and effectively eliminate the effect of uneven illuminations, white keeping the local statistical properties of the processed image the same as the corresponding image under normal lighting condition. Then, the processed images are used for face recognition. Experimental results show that, with the use of PCA for face recognition, the recognition rates can be improved by 46.4%, 40.0%, 8.3% and 37.9% based on the YaleB database, Yale database, AR database and the combined database, respectively, when our proposed algorithm is used.