摘要
针对李春明提出的"无需重新初始化的变分水平集分割模型"存在对内部像素灰度值相近、边缘分离性差、图像分割效果不理想等问题,提出了一种改进的基于核模糊聚类的变分水平集医学图像分割方法.将原始图像进行核模糊C均值聚类处理得到聚类图像,并将其引入初始水平集函数中.然后将改进的边缘指示函数代入李模型中,实现最终的图像分割.通过对人体脑部、肩部MR医学图像进行试验,并采用最大香农熵进行客观评价.结果表明所提出方法的最大香农熵的值在一定程度上大于李模型方法,且运行时间和迭代次数都有所减少,证明了新方法具有良好的分割质量、适应性强,且无需重新初始化.
The existing variational level set without re-initialization model proposed by Li Chunming is less sensitive to images of similar internal pixel gray value with bad edge separation,and the segmentation results are not satisfying.To solve the problems,the variational level set medical image segmentation method was proposed based on kernel fuzzy clustering.The original image was transformed by kernel fuzzy C-means clustering,and the clustering results were introduced into the initial level set function.The improved edge indicator function was brought into the Li model to achieve the ultimate image segmentation.The experiments were conducted on MR images of human brain and shoulder with the proposed method,and the results were objectively evaluated with the maximum Shannon entropy.The experimental results show that the maximum Shannon entropy of the proposed method is higher than that of Li model method to a certain extent,and the proposed method contains less elapsed time and less iteration times at the same time.The proposed method has good segmentation quality and strong adaptability without re-initialization.
出处
《江苏大学学报（自然科学版）》
EI
CAS
CSCD
北大核心
2014年第6期693-698,共6页
Journal of Jiangsu University：Natural Science Edition
基金
国家自然科学基金资助项目(61402204)
江苏省自然科学基金资助项目(BK20130529)
高等学校博士学科点专项科研基金资助项目(20113227110010)
镇江市科技计划项目(SH2014110)
关键词
核模糊C均值聚类
变分水平集
李模型
边缘指示函数
图像分割
kernel fuzzy C-means clustering method
variational level set
Li model
edge indicator function
image segmentation
