Liao Liang, Lin Tusheng, Li Bi, Zhang Weidong
School of Electronic and Information Engineering, South China University of Technology, Guangzhou 510640, China.
Sheng Wu Yi Xue Gong Cheng Xue Za Zhi. 2008 Dec;25(6):1264-70.
A modified algorithm using fuzzy Gibbs random field model and fuzzy c-means (FCM) clustering is proposed for segmentation of Magnetic resonance(MR) brain images. Spatial constraints using the definitions of homogeneity of cliques and fuzzy Gibbs clique potential are introduced in this algorithm. A new modified objective function , which is established by introducing the spatial constraints into the traditional intensity based FCM algorithm, leads to the establishment of new iterative formulas for membership matrix and centroids. This algorithm can improve the performance of corresponding traditional one by modifying the original intensity based segmentation model. Experiments on synthetic images and MR phantoms show the validation of the proposed algorithm, which is usually a better alternative for segmenting medical MR images corrupted by noise.
提出了一种使用模糊吉布斯随机场模型和模糊c均值（FCM）聚类的改进算法，用于磁共振（MR）脑图像的分割。该算法引入了基于团块同质性定义和模糊吉布斯团块势的空间约束。通过将空间约束引入传统的基于强度的FCM算法，建立了一个新的改进目标函数，从而得到了隶属度矩阵和聚类中心的新迭代公式。该算法通过修改原始的基于强度的分割模型，可以提高相应传统算法的性能。在合成图像和MR体模上的实验验证了该算法的有效性，该算法通常是分割受噪声干扰的医学MR图像的更好选择。
