基于谱聚类的多阈值图像分割方法  

Image Segmentation of Multilevel Thresholding Based on Spectral Clustering

作　　者：邹小林[1,2,3] 陈伟福 冯国灿[1,3] 刘志勇[1,3,4] 汤鑫[1,3] 

机构地区：[1]中山大学数学与计算科学学院,广州510275 [2]肇庆学院数学与信息科学学院,肇庆526061 [3]广东省计算科学重点实验室,广州510275 [4]深圳职业技术学院工业中心,深圳518055

出　　处：《计算机科学》2012年第3期246-248,259,共4页Computer Science

基　　金：国家自然科学基金项目(60975083,U0835005)资助

摘　　要：阈值法是图像分割的一种重要方法,在图像处理与目标识别中广为应用。因此,如何确定阈值是图像分割的关键。提出了一种新的图像阈值分割方法,即通过采用新的相似度函数的谱聚类算法(Dcut)确定图像阈值。采用基于灰度级的权值矩阵代替常用的基于图像像素级的权值矩阵描述图像像素的关系,因而算法需要的存储空间及实现的复杂性与其它基于图的图像分割方法相比大大减少。实验表明,该方法分割图像的时间少,且能够单阈值和多阈值分割图像,与现有的阈值分割方法相比,其具有更为优越的分割性能。The thresholding is an important form of image segmentation and is used in many applications that involve image processing and object recognition. Thus, it is crucial to how to acquire a threshold of image segmentation. A novelmultilevel thresholding algorithm was presented in order to improve image segmentation performance at lower computational cost. The proposed algorithm determines the thresholdings by spectral clustering algorithm called Dcut that uses a new similarity function. The weight matrices used in evaluating the graph cuts are based on the gray levels of an image, rather than the commonly used image pixels. For most images, the number of gray levels is much smaller than the number of pixels. Therefore, proposed algorithm occupies much smaller storage space and requires much lower computational costs and implementation complexity than other graph-based image segmentation algorithms. A large number of examples were presented to show the superior performance by using the proposed multilevel thresholding algorithm compared to existing thresholding algorithms.

关 键 词：图像阈值分割 多阈值 谱聚类 Dcut 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]