一种基于LoG算子边缘检测的全局二值化方法

In this paper, several widely used binarization methods are studied and compared, and a new gobal threshold binarization methods that is based on LoG(Laplacian of Gaussian) are proposed. This method aims at image that object and background are not quite separated. Experiments show that this method can get the best binarization threshold and distinguish the object and the background perfectly, meantime, it also can avoid disconnection.