基于LoG算子边缘检测的图像二值化处理

Global threshold binarization method based on LoG algorithms edge detecting

数码管图像的目标和背景分离不明显,直方图分布较复杂。针对该问题,提出基于拉普拉斯高斯(Laplacian of Gaussian,LoG)算子边缘检测的全局二值化方法对其进行处理,该方法通过提取图像边缘部份的像素灰度获得图像二值化的阈值。处理结果表明,与传统的几种方法相比,该方法能够快速选取良好的二值化阈值,较好地区分目标和背景,在相当大模板宽度内图像二值化的结果都令人满意。

As the object and background of numeral characters digital image are not quite separated, this paper proposed a new global threshold binarization methods based on LoG （Laplacian of Gaussian）. This method gets binarization threshold by averaging the gray scale of edge pixels. Experiments show that this method can get the best binarization threshold and distinguish the object and the background perfectly. Perfect binarization image is got at wide range of template width.