The performance of face recognition techniques is greatly challenged by the pose, expression and illumination of the image. For most existing systems, the recognition rate will decrease due to changes in environmental illumination. In this paper, a Homomorphic Wavelet-based Illumination Normalization (HWIN) method is proposed. The purpose of this method is to normalize the uneven illumination of the facial image. The image is analyzed in the logarithm domain with wavelet transform, the approximation coefficients of the image are mapped according to the reference illumination map in order to normalize the distribution of illumination energy resulting from different lighting effects, and the detail components are enhanced to achieve detail information emphasis. Then, a Difference of Gaussian (DoG) filter is also applied to reduce the noise resulting from different lighting effects, which exists on detail components. The proposed methods are implemented on Yale B and Extended Yale B facial databases. The experimental results show that the methods described in this study are capable of effectively eliminating the effects of uneven illumination and of greatly improving the recognition rate, and are therefore more effective than other popular methods.