空间约束和距离改进的 FCM 图像分割算法

摘要

Aiming at the poor noise resistance of FCM when applying to image segmentation,we propose in this paper a fuzzy c-means clustering algorithm which is based on spatial constraints and subspace distance.The algorithm adds a constraint item containing spatial domain information on the basis of original FCMformula,this makes the objective function be minimum when the adjacent pixels tending to the same class in whole.Moreover,it replaces the original Euclidean distance of FCM with the distance between a point and the clustering subspace so as to achieve more accurate clustering effect.Experimental results of artificial image and natural image segmentation show that the proposed algorithm obviously outperforms standard FCMalgorithm and has good antinoise performance.%针对传统 FCM算法在图像分割应用中抗噪性差的问题，提出一种基于空间约束和子空间距离的模糊 C-均值聚类算法。该算法在原 FCM公式的基础上加入一个包含空间领域信息的约束项，使得整体上相邻像素点趋于同一类时，目标函数最小。并将原 FCM的欧氏距离替换为点到聚类子空间的距离，以达到更精准的聚类效果。人造图像和自然图像的分割实验结果表明，该方法明显优于标准的 FCM算法，具有很好的抗噪性能。