Varying illumination conditions affect the appearance of face images significantly. Thus, it severely degrades image-based face recognition performance. This paper presents a novel face image pre-processing approach that deals with the illumination problem to make face recognition robust to illumination variations. In the proposed method, logarithm transform is first used to convert a face image into logarithm domain. Then discrete cosine transform (DCT) coefficients of it are modified to remove illumination variations. The reconstructed log image by inverse DCT of the modified coefficients is used for the final recognition. We achieved 100% face recognition rate on Yale face database B. The proposed method requires no assumption on the light source and any prior information on 3-D face geometry.