A Novel Binarization Approach for License Plate

Abstract

Binarization of a gray scale license plate image is one of the most important steps of license plate recognition (LPR). It segments a license plate image into foreground and background. The foreground contains the characters to be segmented. This paper presents a compensation-based and central-scanned license plate image binarization approach. The approach enhances the license plate image with a contrast-stretching transformation at first. Then it applies a compensation-based binarization technique to the whole image and utilizes a central-scanned method to unify those license plate images with black character and white background after previous binarization. Experimental results on a number of license plate images show that our binarization approach combining compensation and central-scan is efficient even in situations of overexposure and underexposure