Abstract: To make up for the restriction of traditional segmentation methods regarding the number of segmentation and fuzzy clustering method’s lacking in levels of optimization, this paper presents an effective fuzzy segmentation approach based on multiscale linking model for brain MRI. The non-uniformity of gray-scale is corrected after analysis of the bias in brain MRI. A fuzzy inter-scale constraint via antistrophic diffusion linking model is introduced. Two fuzzy distances are developed and embedded into the fuzzy clustering algorithm. Moreover, a multiresolution framework combining the inter-scale and intra-scale constraints is presented. Experimental results show the accuracy and validity of this method in which a large deal of brain MRI data is used.

Key words: multiscale linking model, fuzzy segmentation, bias correction

摘要： 针对传统分割方法在分割数量上的限制，以及模糊聚类方法在层次优化上的不足，提出一种有效的基于多尺度连接模型的人脑磁共振图像模糊分类算法。对脑磁共振图像进行灰度不均匀性校正后，该方法通过非线性扩散连接模型引入尺度间的模糊约束，定义2个模糊距离，并应用到模糊聚类算法中，组合尺度间和尺度内的模糊约束，给出一个多分辨模糊聚类框架。实验结果表明了该算法的准确性和有效性。

关键词: 多尺度连接模型, 模糊分类, 偏场校正

CLC Number: 

• TP391