摘要
针对传统FCM(fuzzy c-means)算法抗噪性差的问题,提出了一种基于一致性分片的模糊c均值聚类算法。为避免额外的空间邻域约束项带来的控制变量设置问题,该算法直接将FCM应用于图像片空间。为减弱空间邻域对图像边缘的模糊,采用基于置信区间的局部多项式交叉近似技术(local polynomial approximation and intersection of confidenec intervals,LPA-ICI)构造自适应形状一致性分片。在脑磁共振图像上的实验表明,与传统的FCM算法相比,该算法具有更高的分割精度和运行效率。
To address the problem that existing FCM algorithms have poor noise immunity, we propose a novel fuzzy c- means clustering (FCM) algorithm based on homogeneous pieces. The algorithm directly applies FCM in the imaging space, so that the control variable setting problem caused by the introduction of additional space neighborhood constraints is avoided. In addition, in order to abate the fuzzy effect to the edge of image by space neighborhood, the local polynomial ap-proximation-intersection of confidence intervals (LPA-ICI) based on confidence interval is adopted to construct the homoge-neous piece with adaptive form. Through the segmentation of brain magnetic resonance imaging, the experimental results show that, compared with the traditional FCM algorithm, this algorithm can make accurate segmentation to the brain mag-netic resonance image. The method is useful and effective for MRI image segmentation.
作者
丁晓峰
何凯霖
DING Xiaofeng HE Kailin(Department of Computing,Chengdu College of University of Electronic Science and Technology of China,Chengdu 611731, P. R. China)
出处
《重庆邮电大学学报(自然科学版)》
CSCD
北大核心
2017年第3期377-381,共5页
Journal of Chongqing University of Posts and Telecommunications(Natural Science Edition)
关键词
模糊C均值聚类
图像分割
一致性分片
脑磁共振图像
fuzzy c-means clustering
image segmentation
homogeneous piece
brain magnetic resonance image