In this paper, we present a new image preprocessing method which is based on Wavelet Transform. Before we extract expression texture information, Wavelet transform will be applied to remove partial high frequency information. And then Gabor Transform will be used to extract expression information. At last we chose the Support Vector Machine to classify facial expressions. Experiments on the Japanese Female Facial Expression (JAFFE) Database illustrate that the Wavelet Transform and Gabor features are effective for facial expression discrimination. Additionally, the experiment results validate the feasibility of this method.