Abstract

A binarization method based on Fuzzy C-Means (FCM) and Human Vision System (HVS) is presented in this paper. Firstly, the image threshold value is obtained according to FCM. Secondly, the standard deviation is used to determine Visual Sensitivity Coefficient (VSC) of each pixel. Finally image binarization is completed according to the gray value, the image threshold and VSC. This paper method can both effectively remove the ghosts and retain the character stroke well. The experimental results show that this paper method is superior to Ostu method and Bernsen method in the threshold value and the entropy. It provides a new effective thought for vehicle license plate image binarization.

Paper Details