计算机工程与应用 ›› 2009, Vol. 45 ›› Issue (30): 165-167.DOI: 10.3778/j.issn.1002-8331.2009.30.050

• 图形、图像、模式识别 • 上一篇    下一篇

一种基于支持向量聚类的图像分割方法

蒋加伏,赵 嘉,胡益红   

  1. 长沙理工大学 计算机与通信工程学院,长沙 410076
  • 收稿日期:2008-06-06 修回日期:2008-09-05 出版日期:2009-10-21 发布日期:2009-10-21
  • 通讯作者: 蒋加伏

Method of image segmentation based on support vector clustering algorithm

JIANG Jia-fu,ZHAO Jia,HU Yi-hong   

  1. Institute of Computer and Communication Engineering,Changsha University of Science and Technology,Changsha 410076,China
  • Received:2008-06-06 Revised:2008-09-05 Online:2009-10-21 Published:2009-10-21
  • Contact: JIANG Jia-fu

摘要: 利用支持向量聚类分类准确、参数少、无监督学习的特点,提出一种基于支持向量聚类的图像分割方法。该方法首先对数据集分块并对每块进行SVC聚类,再取其簇内均值作为K均值聚类样本点,进行聚类,最后将得到的结果进行合并。实验证明该方法不但改变了传统分割方法中人为选取阈值参数的作法,而且受目标和噪声影响小,提高了图像分割的鲁棒性和效果,能够有效地进行图像分割。

关键词: K均值算法, 支持向量聚类算法(SVC), 图像分割

Abstract: Using the benefit of support vector clustering which supplies’ clustering accuracy,few parameters and unsupervised learning,a method of SVC image segmentation is proposed.Firstly,this method divides data sets into some pieces of data block and carries through supporting vector clustering.Secondly,the mean value of per cluster is considered as the sample data of K-means and goes through clustering.At last,the received result is incorporated.Experimental results demonstrate that this method not only has a good result for small targets with more yawp,but also automatically determines threshold values,which are artificial selections in the past.The robustness of image segmentation has been improved and the performance becomes steadier.The proposed methods can segment image effectively.

Key words: K-means algorithm, Support Vector Clustering(SVC) algorithm, image segmentation

中图分类号: