基于图谱理论的FCM图像分割方法研究

Research of FCM for image segmentation based on graph theory

利用图谱理论的思想对传统模糊C-均值(FCM)图像分割方法进行改进——将图谱理论中的权值计算方法引入到FCM方法的距离计算中,较之原来的Euclid距离不仅考虑了各样本空间上的距离,同时考虑了各样本之间的灰度差异,获得更适用于图像分割的模糊隶属度函数,从而得到改进的FCM图像分割方法。通过与传统FCM图像分割方法、基于图谱理论的图像分割方法的实验结果、错分概率及评价指标的对比分析,证明所提出的改进FCM方法能够很好地解决图像分割问题。

Graph theory was utilized to improve image segmentation of traditional Fuzzy C-Means （FCM）. The proposed algorithm used weighting of graph theory to calculate the distance of FCM, compared with Euclid distance, the proposed algorithm not only considered the distance of every sample, but also considered the Grayscale difference of every sample, and gained fuzzy membership function which was suitable for image segmentation. Based on the experimental result, probability of error and index of evaluation through comparing with image segmentation based traditional FCM and image segmentation based graph theory. The improved FCM in this paper is proved to be an appropriate method which is suitable for image segmentation.