结合模糊C均值聚类与图割的图像分割方法  

Image segmentation approach of combining fuzzy clustering and graph cuts

作　　者：王晓飞[1] 郭敏[1] 

机构地区：[1]陕西师范大学计算机科学学院,西安710062

出　　处：《计算机应用》2009年第7期1918-1920,共3页journal of Computer Applications

摘　　要：针对模糊C均值聚类没有考虑像素空间信息的不足,提出一种结合模糊C均值聚类与图割的图像分割方法。以图割理论为基础,考虑到像素的空间信息,建立一个关于标号的全局能量函数,以FCM聚类中心为终端建立多终端网络图,该网络通过α扩展移动算法求解全局最小或近似最小能量函数所对应的标号函数f,在各类间重新划分所有像素点,实现目标正确分割。实验表明,该方法在分割精度、性能、抗噪性等方面均有较大改进。This paper proposed an image segmentation method combining fuzzy C-means clustering and graph cuts. Based on graph cuts, considering the pixel space, establish a global energy function concerning labels, and make the FCM clustering center as terminal to found a multi-terminal network, get the global minimum of the energy function or the approximate minimum of the label function by the α expansion moving algorithm . Re-demarcate all pixels between all classes and get the right targets. Experimental results show that the method has great improvement in accuracy, performance and noise immunity.

关 键 词：图像分割 模糊C均值 聚类 图割 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]