改进的快速模糊C均值聚类的图像分割方法

A modified method for image segmentation with fast fuzzy C-means clustering

传统的模糊C均值(FCM)聚类算法广泛用于图像的自动分割,但该算法没有考虑像素的灰度和空间特征,对噪声和伪斑点图像不可能取得好的分割效果.提出一种改进的算法,在快速的FCM聚类的基础上,运用邻域像素的灰度相似度和聚类分布统计构造新的隶属函数,对图像进行二次聚类分割.该算法具有以下优点:1)有效地抑制了噪声的干扰;2)减少了图像的伪斑点;3)把误分类的像素很容易地纠正过来.对两种类型图像的实验分割结果表明该方法对噪声和伪斑点具有很强的鲁棒性和对像素聚类的正确性.

Conventional fuzzy C-means （FCM） clustering algorithm has been widely used in automated image segmentation. However, it was not successful to segment the noise image and the image with spurious blobs because the gray-scale and spatial characteristics of the pixel were not taken into consideration. Therefore, a modified algorithm was proposed for secondary clustering and segmentation of the image on the basis of fast FCM clustering, and using the gray-scale similarity and cluster distribution statistics of the neighbor pixels to form a new membership function. The advantages of this new method were as follows： 1） it was effective to restrain the noisy interference, 2） it reduced the spurious blobs of the image, and 3） it was ease to correct the misclassified pixels. Experimental results of two types of noisy images indicated that the segmentations were more accurate and robust than those with standard FCM algorithm.