摘要
针对传统模糊C-均值(Fuzzy C-means,FCM)聚类算法对噪声鲁棒性差的问题,提出一种基于空间信息的模糊C-均值噪声图像分割算法。将区域级信息加入FCM目标函数中,并用核度量方法代替传统欧氏距离,计算区域级空间信息与聚类中心的距离,提高算法对噪声的鲁棒性;用原始图像与区域级空间信息的绝对差的倒数和其本身约束原始图像和区域信息项,实现约束项参数的自适应选择;利用连通分量滤波,消除聚类结果中出现的过分割现象,提高分割精度。含噪合成图像和彩色图像实验表明,所提算法在模糊分割系数、模糊分割熵、分割精确度、平均交互比和归一化互信息等方面均优于其他几种聚类算法。
To deal with the problem of poor robustness of traditional Fuzzy C-means(FCM)clustering algorithm to noise,a fuzzy C-means algorithm for noisy image segmentation based on spatial information is proposed.Firstly,regional-level information is added to the FCM objective function,the kernel measurement method is used to replace the traditional Euclidean distance,and the distance between the regional-level spatial information and the cluster center is calculated,thus improving the algorithm's robustness to noise.Secondly,the reciprocal of the absolute difference between the original image and regional spatial information is used to constrain the original image and regional information,and the adaptive selection of constraint parameters is realized.Finally,the connected component filtering is used to eliminate the over-segmentation phenomenon in the clustering results and improve the segmentation accuracy.Experiments on noisy synthetic images and color images show that the proposed algorithm is superior to other clustering algorithms in terms of fuzzy segmentation coefficient,fuzzy segmentation entropy,segmentation accuracy,average interaction ratio and normalized mutual information.
作者
李力
陈息坤
LI Li;CHEN Xikun(School of Internet of Things Engineering,Guangdong Polytechnic of Science and Technology,Guangzhou 510640,China;School of Mechatronic Engineering and Automation,Shanghai University,Shanghai 200444,China)
出处
《无线电工程》
北大核心
2023年第10期2295-2302,共8页
Radio Engineering
基金
广东省普通高校工程技术中心“智能装备制造工程技术研究中心”(2021GCZX018)。
关键词
噪声图像分割
模糊C-均值聚类
区域级信息约束
核度量方法
连通分量滤波
noisy image segmentation
fuzzy C-means clustering
region-level information constraints
kernel measurement method
connected-component filtering
