计算机工程与应用 ›› 2009, Vol. 45 ›› Issue (21): 99-102.DOI: 10.3778/j.issn.1002-8331.2009.21.029

• 图形、图像、模式识别 • 上一篇    下一篇

结合相关反馈日志与语义网络的图像标注

杨 珺1,2,王继成2,邢丹俊2   

  1. 1.上海电力学院 计算机信息与工程学院,上海 200090
    2.同济大学 电子与信息工程学院,上海 201804
  • 收稿日期:2008-04-25 修回日期:2008-07-21 出版日期:2009-07-21 发布日期:2009-07-21
  • 通讯作者: 杨 珺

Using relevance feedback log and semantic network for image annotation

YANG Jun 1,2,WANG Ji-cheng2,XING Dan-jun2   

  1. 1.Shanghai University of Electric Power,Shanghai 200090,China
    2.School of Electronics and Information Engineering,Tongji University,Shanghai 201804,China

  • Received:2008-04-25 Revised:2008-07-21 Online:2009-07-21 Published:2009-07-21
  • Contact: YANG Jun

摘要: 连接高层语义和低层视觉特征的图像语义标注技术能够很好地表示图像的语义,提出并实现了一种结合相关反馈日志与语义网络的图像标注方法。该方法以收集的用户相关反馈日志为基础获得图像的语义信息,通过计算图像间的语义相似度进行语义聚类并采用语义传播的方式实现图像的语义标注。实验结果表明,随着相关反馈日志库的不断扩充,图像库中越来越多的图像会在反馈的过程中得到标注且标注的准确率会随着反馈次数的增加而趋于稳定。

Abstract: Image annotation,a technique for connecting image semantics and visual features,can be used to present image semantics well.A method for image annotation using relevance feedback log and semantic network is presented.Firstly,the image semantics are acquired by users’ relevance feedback log;then semantic clustering is carried out based on semantic similarity;lastly automatic image annotation is realized through semantic propagation.The experimental results indicate that increasing images will be annotated as increasing relevance feedback log and the annotation accuracy will be stable with the increment of relevance feedback.