基于Nystrom方法的水平集医学图像分割算法  

Level set method of medical image segmentation algorithm based on Nystrom method

作　　者：王晓丹[1] 张龙波[1] 王雷[1] 刘晨 WANG Xiao-dan;ZHANG Long-bo;WANG Lei;LIU Chen(School of Computer Science and Technology,Shandong University of Technology,Zibo 255049,China)

机构地区：[1]山东理工大学计算机科学与技术学院,山东淄博255049

出　　处：《山东理工大学学报（自然科学版）》2018年第3期13-16,20,共5页Journal of Shandong University of Technology:Natural Science Edition

基　　金：国家自然科学基金青年科学基金项目(61502282);山东省自然科学基金青年科学基金项目(ZR2015FQ005);山东理工大学博士科研启动经费资助项目(414023);淄博市校城融合发展计划项目(2016ZBXC049)

摘　　要：现有的基于最小化区域扩展拟合能量的图像分割模型,对于边缘模糊、噪声强的图像存在易产生边缘泄露的现象,导致分割效果不理想.针对这种现象提出了一种基于Nystrom方法的水平集医学图像分割算法.算法将原始图像通过Nystrom方法采样,近似估算相似矩阵和特征向量,通过k-means算法将特征向量聚类,最后利用水平集分割方法实现图像分割.实验结果表明,与基于最小化区域扩展拟合能量的图像分割模型相比,在相同的迭代次数中,分割时间减少,相似度系数提高.The existing image segmentation model based on minimized region-extended fitting energy has a tendency to produce edge leakage for edges with fuzzy and strong noise,resulting in dissatisfactory segmentation.A new level set of medical image segmentation algorithm based on Nystrom method is proposed for this phenomenon.Firstly,the original image is sampled by the Nystrom method,and the similarity matrix and eigenvector are estimated approximately.Then the feature vector is clustered by the k-means algorithm.Finally,the image segmentation is realized by the level set segmentation method.The experimental results show that the segmentation time is reduced and the similarity coefficient is improved with the same number of iterations compared with the minimized region-extended fitting energy for image segmentation.

关 键 词：水平集 图像分割 聚类 NYSTROM方法 相似矩阵 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]