结合空间信息的FCM脑图像分割

针对传统模糊C均值聚类算法(FCM)在图像分割时没有利用图像的空间信息而对噪声敏感、分割结果不够准确的问题,提出一种结合空间信息的FCM改进算法.该算法利用邻域像素间的灰度差异计算出邻域加权系数,并利用该系数对中心像素的隶属度进行更新,控制邻域像素对中心像素的不同影响;该算法还利用了快速FCM算法对图像进行初始分割.对MRI脑图像分割的实验结果表明FCM改进算法简单有效,具有较强的抗噪能力,能取得较好的图像分割结果.

The Brain Image Segmentation Based on FCM with Spatial Information

The standard fuzzy C-means （FCM） algorithm is sensitive to noise and can＇t get exact results, because of not taking into account the spatial information of image. To overcome these drawbacks, an improved FCM algorithm using neighborhood information is proposed in this paper. Using the gray differences of neighborhood pixels, the proposed algorithm computes the neighborhood weighted coefficient and modifies the membership of central pixel with the coefficient, controlling the effect of neighborhood pixels on central pixel. In addition, a fast FCM algorithm is used at the beginning of segmentation. The experimental results of MRI brain image segmentation show that the improved FCM algorithm is simple and efficient, has stronger anti-noise property and can give finer results of image segmentation.