模糊C均值聚类与硬聚类的性能比较及改进

本文以灰度值的图像分割为基础,对模糊C均值聚类算法(Fuzzy C-means,FCM)1]和硬聚类进行了详尽的讨论,在此基础上对两者进行了比较,包括两者的迭代速度比较和两者的分割效果比较,聚类中心的初始化对迭代速度和分割效果的影响,并以此为基础对FCM聚类算法进行了改进。实验表明,改进的FCM聚类算法在迭代速度和分割效果方面都明显优于原始的FCM聚类算法。

Fuzzy C-means' Performance Comparison with Hard Clustering And Improvement

Based on gray scale image segmentation, this paper evaluates the performance of Fuzzy C-means Clustering and Hard Clustering algorithm. The iterative speed, segmentation quality and how the initialization value of clustering center affects speed and quality is being discussed in detail. According to the results, the FCM clustering is improved to get better speed and quality, which is being proved to be true.