结合类的内聚性的Otsu阈值计算方法研究

A Novel Thresholding Approach Devised by Combining Class Cohesiveness into Otsu Method

Otsu算法是一种传统的阈值计算方法,并在实践中得到了广泛的应用。然而它在前景与背景分界较为模糊的图像上效果不是很理想。本文提出了一种对Ostu方法进行改进的阈值计算方法,该方法在将图像进行均衡化处理后,将前景和背景像素类的内聚性结合到Otsu方法中,并给出了一个改进的分类判别函数。

Otsu method is a traditional thresholding method, which has been widely applied to practice. However, it shows drawbacks when dealing with images with diffused foreground and background. This paper proposes an approach to improving Otsu method, which combines elass eohesiveness into standard Otsu on histogram equilized images forming a discriminant function. Experimental results show that the improved method performs well eompared to standard Otsu.