计算机应用 ›› 2010, Vol. 30 ›› Issue (06): 1587-1589.

• 图形图像处理与模式识别 • 上一篇    下一篇

基于最小割的极化特征图像分割

史彩云1,林伟2,李 旭2,温金环2   

  1. 1. 西北工业大学
    2.
  • 收稿日期:2009-11-30 修回日期:2010-01-21 发布日期:2010-06-01 出版日期:2010-06-01
  • 通讯作者: 史彩云
  • 基金资助:
    国家自然科学基金资助项目;西北工业大学科技创新基金资助项目

Polarized characteristics image segmentation based on minimum cut

  • Received:2009-11-30 Revised:2010-01-21 Online:2010-06-01 Published:2010-06-01

摘要: 针对极化合成孔径雷达(SAR)所固有的斑点噪声很难分割出精确结果的问题,提出了一种基于图论的极化SAR图像分割方法。该方法结合极化SAR的多个极化特征, 用K均值聚类算法得到像素的初始标号,然后建立一个关于标号的能量函数并构造相应的网络,用最小割方法求取网络中全局能量函数的近似最优解,由此得到每个像素点的恰当标号,最终完成图像的准确分类。该方法与传统的分割方法相比,能够充分考虑极化SAR图像的全局信息和极化特征对图像进行精确的分割。实验结果证明,该算法具有较好的分割效果。

关键词: 极化SAR, 图像分割, 最小割, K-Means聚类算法, 能量函数

Abstract: It is difficult to get accurate segmentation results because of inherent speckle noise of Polarimetric Synthetic Aperture Radar (POL-SAR); therefore, a graph-based POL-SAR image segmentation method was presented. The method used K-Means clustering algorithm to get the initial label of every pixel with reference to the combined multiple polarized characteristics. Then we used minimum cut method to get the optimal solution of global energy function approximately in these networks after establishing a labeled energy function and constructing the corresponding networks. Therefore, we could get the proper label of every pixel and complete the correct classification. Compared with other traditional segmentation methods, this method considers the global information and polarized characteristics of POL-SAR image adequately so as to get the accurate segmentation. The experimental results show the algorithm has better segmentation effect.

Key words: polarimetric synthetic aperture radar(POL-SAR), image segmentation, Minimum cut, K-Means clustering algorithm, energy function