基于邻域信息约束的模糊聚类图像分割方法  

The method of image segmentation with fuzzy cluster based on neighborhood information constraint

作　　者：常晓茹 郑寇全 李娜 王萌 CHANG Xiao-ru ZHENG Kou-quan LI Na WANG Meng(Information Section of Lintong Sanatorium, PLA , Xi＇an 710600, China Department of Information Service, Xi ＇ an Communication Institue , Xi＇an 710106, China Unit 93413, PLA , Yongji 044500, China)

机构地区：[1]解放军临潼疗养院信息科 [2]西安通信学院信息服务系 [3]中国人民解放军93413部队

出　　处：《电子设计工程》2017年第21期6-10,共5页Electronic Design Engineering

基　　金：国家自然科学基金面上项目(61272011);国家自然科学基金青年项目(61309022);陕西省自然科学基金青年项目(2013JQ8031)

摘　　要：针对基于模糊C均值聚类(FCM)的图像分割算法仅利用像素的灰度信息、噪声抑制不理想、算法鲁棒性不高的问题,提出了一种基于像素邻域信息约束的FCM图像分割算法。该算法在模糊目标函数中引入邻域信息约束,通过约束系数自适应调节控制邻域信息约束强度,自优化迭代更新聚类中心和聚类隶属度矩阵,使模糊目标函数收敛到最小,并利用像素最优聚类隶属度去模糊化操作实现图像分割。实验结果表明,该算法在获得较高的图像分割精度的同时,具有较强的噪声抑制能力。According to the problem of the fuzzy C-means clustering （FCM） algorithm for using only gray information in the image segmentation, the bad performance of noise suppression and robustness, the FCM algorithm for image segmentation based on pixel neighborhood constraint information is proposed. By introducing the neighborhood information constraint into fuzzy objective function, the neighborhood information constraint intensity is controlled from the constraint coefficient adaptive, the clustering center and membership matrix are iterative update self-adaption, the fuzzy objective function converges to the minimum. The results of image segmentation are obtained by defuzzification for optimal pixel membership. Finally, the classical instance shows that the algorithm has higher accuracy of image segmentation and noise suppression.

关 键 词：邻域信息 约束系数 模糊C均值 图像分割 

分 类 号：TN957.52[电子电信—信号与信息处理]