Tilt correction is a very important link in the vehicle license plate automatic recognition process. By Principal Component Analysis (PCA), the image character coordinates are arranged into two-dimension covariance matrix, which centralization is operated. Then the feature vector and the rotation angle α are computed. The whole image is rotated by α and image horizontal tilt correction is performed. In the vertical tilt correction course, two correction methods, namely, PCA method and the Line Fitting Method Using K-means Clustering (LFMUKC) are proposed to compute the vertical tilt angle θ. The rotated image is done to Shear Transform and the final correction image is obtained. The experimental results show that, this paper approach can be implemented easily, also can quickly and accurately get the tilt angle. It provides a new effective way for tilt correction of vehicle license plate.