基于非局部信息和子空间的模糊C有序均值聚类的图像分割算法  

Image Segmentation Algorithm Based on Non-Local Information and Subspace for Fuzzy C-Ordered Mean Clustering

作　　者：陈阳 黄成泉 覃小素 彭家磊 雷欢 周丽华 Chen Yang;Huang Chengquan;Qin Xiaosu;Peng Jialei;Lei Huan;Zhou Lihua(School of Data Science and Information Engineering,Guizhou Minzu University,Guiyang 550025;Engineering Training Center,Guizhou Minzu University,Guiyang 550025)

机构地区：[1]贵州民族大学数据科学与信息工程学院,贵阳550025 [2]贵州民族大学工程技术人才实践训练中心,贵阳550025

出　　处：《计算机辅助设计与图形学学报》2025年第3期506-518,共13页Journal of Computer-Aided Design & Computer Graphics

基　　金：国家自然科学基金(62062024);贵州省省级科技计划(黔科合基础-ZK[2021]一般342);贵州省研究生教育教学改革重点项目(黔教合YJSJGKT[2021]018);贵州省教育厅自然科学研究项目(黔教技[2022]015).

摘　　要：针对模糊C有序均值聚类算法没有考虑图像空间信息,导致难以有效地分割含噪图像的问题,提出一种基于非局部信息和子空间的模糊C有序均值聚类(non-local information and subspace for fuzzy C-ordered means,SFCOM-NLS)算法.首先,利用图像中给定的相似邻域结构的像素提取当前像素的非局部空间信息;其次,计算每个像素的典型性,并对其进行排序,在每次迭代中更新像素的典型性,提高像素聚类的准确性,解决在聚类过程中存在相似类导致的误分类问题;最后,引入子空间聚类概念,为图像不同维度分配适当的权重,提高彩色图像的分割性能.在含噪合成图像和公开数据集BSDS500,MSRA100和AID上实验结果表明,所提算法的模糊划分系数、模糊划分熵、分割精度和标准化互信息平均值分别达到了95.00%,6.66%,98.77%和95.54%,均优于对比的同类算法.Aiming at the problem that the fuzzy C-ordered means algorithm does not consider the spatial information of the image,which makes it difficult to effectively segment the noisy image,a fuzzy C-ordered means clustering algorithm based on non-local information and subspace is proposed.Firstly,the non-local spatial information of the current pixel is extracted by using the pixels of the given similar neighborhood structure in the image.Secondly,the typicality of each pixel is calculated and sorted,and in each iteration update the typicality of pixels and to solve the problem of misclassification caused by similar classes during the clustering process.Finally,the concept of subspace clustering is introduced to allocate appropriate weights to different dimensions of the image,thus improving the segmentation performance of color images.The experimental results on the noisy synthetic images and public datasets BSDS500,MSRA100 and AID show that the partition coefficient,partition entropy,segmentation accuracy and normalized mutual information average of the proposed algorithm achieves 95.00%,6.66%,98.77%and 95.54%,respectively,which are better than similar methods in comparison.

关 键 词：非局部空间信息 子空间聚类 模糊C有序均值聚类 噪声图像分割 鲁棒性 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]