摘要
给出一种基于图形模糊聚类(fuzzy clustering method on picture fuzzy sets,PFCM)的改进鲁棒分割算法。该算法将样本聚类所对应的中立度和拒绝度相结合,构造幂积型表达式,将该表达式作为正则项嵌入聚类目标函数,通过目标函数最小化存在极值的必要条件获得改进的图形模糊聚类迭代方法。再将邻域像素灰度信息嵌入改进的图形模糊聚类目标函数,利用拉格朗日乘子法获得图像分割的像素聚类迭代算法。通过标准图像及噪声干扰的分割测试,结果表明,与模糊C-均值聚类、直觉模糊聚类算法和图形模糊聚类分割算法相比,改进算法对无噪图像分割更有效;与鲁棒模糊C-均值聚类和鲁棒直觉模糊聚类算法相比,改进算法对噪声图像分割具有更强的抗噪能力。
A modified picture fuzzy clustering and its image segmentation algorithm are proposed.In this algorithm,based on existing picture fuzzy clustering,the regularization term combining neutral degree with refused degree is firstly embedded into its objective function,a new picture fuzzy clustering iteration method is then obtained by the necessary condition of the existent value of the minimization of object function.Secondly,the neighborhood information of current pixel clustered is embedded into the improved picture fuzzy clustering objective function to enhance the robust ability of its image segmentation,and then a new iterated picture fuzzy clustering formulations for image segmentation is deduced by the Lagrange multiplier method.Segmentation results of standard gray images interrupted by noise show that the proposed picture fuzzy clustering algorithm is more effective than those of fuzzy c-means clustering algorithm and intuitionistic fuzzy clustering algorithm,and has stronger ability against noise than those robustness of fuzzy c-means clustering and robustness of intuitionistic fuzzy clustering algorithm.
出处
《西安邮电大学学报》
2017年第5期37-43,共7页
Journal of Xi’an University of Posts and Telecommunications
基金
国家自然科学基金资助项目(61671377)
陕西省自然科学基金资助项目(2014JM8331
2014JQ5183
2014JM8307)
陕西省教育厅科学研究计划资助项目(2015JK1654)
关键词
图形模糊聚类
图像分割
聚类有效性
分割鲁棒性
picture fuzzy clustering,image segmentation,clustering effectiveness,segmentation robustness
