摘要：

模糊C均值算法是图像分割的常用方法，但该算法对噪声非常敏感。为此，提出一种新算法，在模糊C均值算法基础上引进Type-2模糊理论，以提高算法的分割准确性和鲁棒性。该算法对模糊C均值算法中每一个样本的隶属度进行分段线性拉伸，利用拉伸的结果作为一个新的隶属度函数，并用该函数对图像进行分割。实验结果表明，该算法准确性较高，且具有良好的抗噪能力。

关键词: 图像分析, 图像分割, 模糊聚类, 二型模糊, 隶属函数

Abstract:

The Fuzzy C-Means(FCM) algorithm is one of the most popular image segmentation methods, but the FCM is sensitive to noise. A new image segmentation algorithm is proposed aiming to improve the segmentation precision and robustness of the FCM algorithm by introducing the Type-2 fuzzy theory. A piecewise-linear stretching method is applied to the membership values for each pixel. These membership values are derived using the FCM algorithm. The result of stretching defines a new membership function, which is used for image segmentation. Experimental results show the algorithm has higher image segmentation accuracy and better noise immune ability.

Key words: image analysis, image segmentation, fuzzy clustering, Type-2 fuzzy, membership function

中图分类号: 

• TP311.52