一种模糊C均值图像分割改进算法  

An improved fuzzy C-means image segmentation algorithm

作　　者：刘景康 刘健[4] 赵坤[1,2] 张浩[1,2] LIU Jing-kang;LIU Jian;ZHAO Kun;ZHANG Hao(Beijing Key Laboratory of Communication and Radio Frequency IC Technology for Next Generation,Institute of Microelectronics of Chinese Academy of Sciences,Beijing 100029,China;Institute of Microelectronics of Chinese Academy of Sciences,Beijing 100029,China;University of ChineseAcademy of Sciences,Beijing 100049,China;School of Computer & Communication Engineering,University of Science & Technology,Beijing 100083,China)

机构地区：[1]中国科学院微电子研究所新一代通信射频芯片技术北京市重点实验室,北京100029 [2]中国科学院微电子研究所,北京100029 [3]中国科学院大学,北京100049 [4]北京科技大学计算机与通信工程学院,北京100083

出　　处：《电子设计工程》2019年第3期6-10,15,共6页Electronic Design Engineering

基　　金：国家科技重大专项(2015ZX03001013-002);湖北省中国科学院科技合作专项项目资助

摘　　要：提出了一种基于空间邻域信息和灰度分布信息的模糊C均值图像分割改进算法。将空间邻域信息引入目标函数中,充分考虑邻域像素对聚类的影响,定义新的特征距离为空间距离和灰度距离的加权和,并根据相同灰度的像素点在各个聚类集合的分布情况对特征距离进行修正,建立新的目标函数实现图像分割。将该算法应用于医学尿沉渣图像的分割中,实验结果表明,相对于FCM_S算法改进的FCM算法的分割质量更好,能够有效抑制噪声的干扰,鲁棒性较强,满足尿沉渣图像分割的应用要求。An improved fuzzy C means image segmentation algorithm based on local information and grayscale distribution is proposed.Introduced local information into objective function,defined characteristic distance as weighted combination of spatial distance and grayscale distance,fully considered local pixels’influence to clustering.Amend the characteristic distance with the distribution of pixels with same grayscale in every cluster set,established a novel objective function to implement image segmentation.Applicate this algorithm in the segmentation of medical urinary sediment images,experimental results indicate that improved algorithm achieved better segmentation quality than FCM_S algorithm and suppressed noise effectively.This algorithm is quite robust,can meet the application requirement of urinary sediment image segmentation.

关 键 词：模糊C均值 图像分割 邻域信息 灰度分布 尿沉渣图像 

分 类 号：TP751.1[自动化与计算机技术—检测技术与自动化装置]