Abstract: The threshold segmentation is widely used in image segmentation，but the traditional single-threshold segmentation method has some shortcomings.This paper applies fuzzy technologies to traditional Otsu segmentation，considering the characteristics of human visual system and the uncertainty of fuzzy images.It divides multi-regional with a fast multi-threshold segmentation method，introduces degree of membership to Otsu to find optimal segmentation threshold.The final segmented result is used to segment images.The result of the simulated experiment proves that the method can improve the situations of fuzzy，multi-objective and incomplete segmentation caused by noise，light and other interference factors.The method can provide satisfactory segmented result.

Key words: multi-threshold dividing, degree of membership, maximum between-class variance

摘要： 阈值分割是一种使用较广的图像分割处理方法，传统的单阈值分割方法存在一定的缺陷。针对人眼视觉系统的特点并考虑图像的模糊不确定因素，将模糊技术应用到传统的最大类间方差分割法。用一种快速多阈值分割方法进行多区域划分，引入隶属度改进最大类间方差法，寻找区域最优分割阈值，进而分割图像。仿真实验结果表明：用该方法分割图像可以大大改进因噪声、光照或其他干扰因素造成的模糊、多目标、分割不完全的情况，分割效果明显改善。

关键词: 多阈值分割, 隶属度, 最大类间方差