改进的模糊C均值聚类算法

Improved Method for Fuzzy C-Means Clustering Algorithm

由于传统的模糊C均值(fuzzy C-means,FCM)算法没有考虑像素点的空间邻域信息,仅涉及像素的单点灰度,在处理含有噪声的图像时有很大的局限性,因此分割效果较差。针对FCM的缺陷,提出一种新的改进算法,该算法引入Gibbs随机场,将Gibbs随机场先验概率与像素点隶属度的乘积作为新的像素隶属度。实验表明,改进后的算法有良好的分割效果,既可以较为完整地保留图像边界细节,又能较好地去除图像的噪声。

The traditional fuzzy C-means （FCM） algorithm has great limitations in dealing with the noisy images owing to not considering the spatial information of the pixels and only involving the pixel gray of a single point, so it＇s poor in segmenting an image. For the defects of the FCM algorithm, a new improved algorithm is proposed in this article, in which a product of Gibbs priori probability and the membership is regarded as the new pixel membership. Experimental results show that the improved algorithm has a good segmentation result; it can retain more complete edge details of image and can remove the image noise more effectively.