引用本文:吴嘉昕,王小鹏,焦建军,陈浩然.增强空间信息的快速自适应模糊聚类图像分割算法[J].控制理论与应用,2026,43(3):480~490.[点击复制]
WU Jia-xin,WANG Xiao-peng,JIAO Jian-jun,CHEN Hao-ran.Fast adaptive fuzzy clustering image segmentation algorithm with enhanced spatial information[J].Control Theory & Applications,2026,43(3):480~490.[点击复制]
增强空间信息的快速自适应模糊聚类图像分割算法
Fast adaptive fuzzy clustering image segmentation algorithm with enhanced spatial information
摘要点击 607  全文点击 97  投稿时间：2023-12-22  修订日期：2025-12-22
查看全文  查看/发表评论  下载PDF阅读器   HTML
DOI编号  10.7641/CTA.2024.30821
  2026,43(3):480-490
中文关键词  图像分割  模糊C-均值聚类  增强空间信息  稀疏正则化  自适应参数
英文关键词  image segmentation  fuzzy C-means clustering  enhanced spatial information  sparse regularization  adap tive parameters
基金项目  国家自然科学基金(61761027)，兰州市科技计划项目(2023-3-104)，甘肃省优秀研究生“创新之星”项目(2023CXZX–510)
作者单位E-mail
吴嘉昕 兰州交通大学电子与信息工程学院 1003637459@qq.com 
王小鹏* 兰州交通大学电子与信息工程学院 wangxiaopeng@mail.lzjtu.cn 
焦建军 兰州交通大学电子与信息工程学院  
陈浩然 兰州交通大学电子与信息工程学院  
中文摘要
      针对模糊C-均值聚类(FCM)算法对图像空间信息利用不充分,导致对噪声鲁棒性差以及目标函数自适应 参数设置较为复杂的问题,本文提出了一种增强空间信息的快速自适应模糊聚类图像分割算法.首先,本文定义了 一种新的增强空间信息运算,该运算利用调和系数将图像的局部信息和非局部空间信息相结合并整合到FCM聚类 中,以提高算法的鲁棒性;其次,提出一种快速自适应参数设置的方法,并为原始图像与增强空间信息分配更为高效 的自适用参数,进而实现目标函数的关键参数快速自适应计算;最后,将稀疏正则化引入FCM目标函数,减少了算 法的运行时间.此外,所提出算法还设计了一个3步迭代算法用于求解基于稀疏正则化的FCM模型,该算法由拉格 朗日乘子、硬阈值算子和归一化算子构成.在合成图像和不同数据集上的真实图像实验表明,所提出算法在模拟噪 声的条件下,其分割性能和运算效率优于其他同类型的算法.
英文摘要
      Addressing the inadequate utilization of image spatial information by the fuzzy C-means (FCM) algorithm, leading to poor robustness against noise and increased complexity in the adaptive parameter setting for the objective func tion, a rapid adaptive fuzzy clustering image segmentation algorithm enhancing spatial information is proposed. Firstly, a novel operation based on enhanced spatial information is defined. This operation combines local and non-local spatial information of the image into the FCM clustering using harmonic coefficients to enhance the algorithm’s robustness. Sec ondly, a method for rapid adaptive parameter setting is proposed, which efficiently assigns adaptive weights to the original image and enhanced spatial information. This will enable the swift adaptive computation of key parameters for the ob jective function. Finally, sparse regularization is introduced into the FCM objective function, reduced algorithm runtime. Additionally, a three-step iterative algorithm is designed to solve the sparse regularized FCM model, composed of Lagrange multipliers, hard threshold operators, and normalization operators. Experiments on synthetic images and real images from various datasets demonstrate that the proposed algorithm exhibits superior segmentation performance and computational efficiency compared to other algorithms of similar nature under simulated noise conditions.