Vehicle license image segmentation using wavelet transform

In this paper, a new multiscale method of vehicle license image segmentation based on wavelet transform is proposed. This analysis utilizes the local wavelet transform modulus maxima as the image edge at the multiple scales, and combines the multiscale edge information. Then a template matching method is applied to segment the vehicle license image based on edge density analysis and character edge spatial feature after eliminating the long straight lines noise. The approach integrates multiple scale edge information and overcomes the shortcoming of traditional single scale analysis. This advantage has special significance for the hazy image. The experiment with about three hundred images obtained from natural environment shows that the performance of this approach is better than the traditional one, especially for the hazy images.