Abstract: During the research for the license plate recognition based on the Digital Image Processing Technology, because the location and segmentation of license plate are usually based on the binarized image, the image binaryzation is a key step, and the results of the image binaryzation will impact the location and segmentation directly. The most important part of the image binaryzation is the determination of threshold. In this paper, several methods have been compared, and also the basic principles and computer simulation results of inter-class varimax and maximum entropy method have been analyzed. But we have found that these methods are not optional when used to process vehicle license plate images. Therefore, a new method has been put forward, which is the arithmetic of license plate image binarization based on image histogram. Experiments have proven that this arithmetic is optimal when used to process vehicle license plate images.

Key words: license plate recognition, binaryzation, threshold, interclass varimax, maximum entropy method, image histogram

摘要： 在基于数字图像处理的车牌识别技术的研究过程中，因为车牌定位和字符分割一般都是基于车牌区域的二值化结果进行的，因此对图像的二值化是非常关键的一步，二值化的效果直接影响到后面的车牌定位以及字符分割。二值化最主要的是阈值的确定，文章对比了几种常规的车牌图像全局阈值二值化处理算法，并重点分析了最大类间方差法和KSW 最大熵法的基本原理和微机仿真结果，发现用于车辆图片是这些方法不够理想，然后提出一种新的方法—基于图像直方图波形分析的车牌图像二值化处理算法，经过大量试验证明该算法在进行车牌图像二值化处理时效果非常理想。

关键词: 车牌识别, 二值化, 阈值, 最大类间方差法, 最大熵法, 图像直方图