Hendra Kusuma, Wirawan, Adi Soeprijanto : Gabor Phase-based Face Recognition Under Varying Illumination Using Supervised-PCA
Ir. Hendra Kusuma M.Eng.Sc.
Prof.Dr.Ir Adi Soeprijanto M.T.
Recognizing faces with illumination variations is a challenging task since the performance of face recognition can be degraded by Illumination variations that occur on face images. In this paper we propose a novel approach for handling illumination variations for face recognition. We use Gabor-phase features which invariant to changes in intensity or contrast and then we apply Supervised PCA (Principal Component Analysis) to Gabor feature vectors for dimension reduction and also for class separability enhancement. A database of 640 single light source images of 10 individuals with 64 different illumination conditions from the Yale B face database is used to test the method. Experimental results demonstrate that the proposed face recognition method is robust to illumination variation and can achieves 98% recognition rate when the the number of eigenvectors-used is 60%.