Vehicle license plate localization plays an important role in vehicle license plate recognition system. Actually, there are many possible areas similar to license plate within an image. In order to facilitate the accuracy of vehicle license plate recognition, it is important to localize accurately the area of vehicle license plate for obtaining useful information from an image. In this study, a two-step approach for vehicle license plate localization is proposed. The first step is to perform spectrum analysis on the image for extracting features. Then, wavelet transform and support vector machines is used to localize the area of vehicle license plate. The proposed method have been evaluated through 300 test cases, and the correctness of license plate localization is 96.6 %. Thus, integrating the proposed method into license plate recognition system will advance its correctness.