基于聚类与局部统计的水平集医学图像分割  

Level set segmentation based on clustering and local statistics for medical image

作　　者：蒋颖[1] 覃晓[1] 元昌安[1] 刘致锦 JIANG Ying QIN Xiao YUAN Chang-an LIU Zhi-jin(School of Computer and Information Engineering, Guangxi Teachers Education University, Nanning 530023, Chin)

机构地区：[1]广西师范学院计算机与信息工程学院,广西南宁530023

出　　处：《计算机工程与设计》2016年第11期3058-3062,3079,共6页Computer Engineering and Design

基　　金：国家自然科学基金项目(61363037);南宁市邕宁区科学研究与技术开发计划基金项目(20150328A)

摘　　要：为增强水平集主动轮廓算法的噪声鲁棒性,考虑医学图像的灰度复杂、拓扑结构多变等情况,发挥水平集主动轮廓模型不依赖于初始轮廓的特性,提出一种基于聚类与局部统计的水平集医学图像分割方法。对传统的K均值与RSF模型进行改进,提高噪声的鲁棒性,能有效地处理真实医学图像的灰度非均匀性。进行CT和MR医学实验验证,实验结果表明,该方法能够比传统方法更有效地应对噪声干扰,提高了医学图像的分割精度和效果。To enhance the noise robustness of level set active contour algorithm,considering the medical image gray-scale complex,changing topological structure,and so on and so forth,using the feature that level set active contour model is not dependent on the characteristics of the initial contour,level set segmentation based on clustering and local statistics for medical image was proposed.The methods of traditional K-means and RSF model were improved,effectively improving the noise robustness,and the intensity inhomogeneity was effectively dealt with,and CT and MR medical experimental verification was carried out.Experimental results show that the method is more effective than the traditional method when dealing with noise,the precision and effects of medical image segmentation were improved.

关 键 词：水平集主动轮廓 聚类算法 局部鲁棒统计信息 高斯白噪声 医学图像分割 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]