计算机工程与应用 ›› 2010, Vol. 46 ›› Issue (24): 193-196.DOI: 10.3778/j.issn.1002-8331.2010.24.058

• 图形、图像、模式识别 • 上一篇    下一篇

一种基于SVM和主动学习的图像检索方法

张玉芳1,陈 卓1,熊忠阳1,刘 君1,2,王银辉1   

  1. 1.重庆大学 计算机学院,重庆 400044
    2.重庆市广播电视大学 理工学院,重庆 400052

  • 收稿日期:2009-05-25 修回日期:2009-08-03 出版日期:2010-08-21 发布日期:2010-08-21
  • 通讯作者: 张玉芳

Image retrieval method based on SVMs and active learning

ZHANG Yu-fang1,CHEN Zhuo1,XIONG Zhong-yang1,LIU Jun1,2,WANG Yin-hui1   

  1. 1.College of Computer Science,Chongqing University,Chongqing 400044,China
    2.College of Engineering Science,Chongqing Radio & Television University,Chongqing 400052,China

  • Received:2009-05-25 Revised:2009-08-03 Online:2010-08-21 Published:2010-08-21
  • Contact: ZHANG Yu-fang

摘要: 主动学习已被证明是提升基于内容图像检索性能的一种重要技术。而相关反馈技术可以有效地减少用户标注。提出一种主动学习算法,带权Co-ASVM,用于改进相关反馈中样本选择的性能。颜色和纹理可以认为是一张图片的两个充分不相关的视图,分别计算颜色和纹理两种特征空间的权值,并在两种特征空间上分别进行SVM学习,对未标注样本进行分类;为了减少反馈样本的冗余,提出一种K-means聚类的主动反馈策略,将未标注样本返回给用户标注。实验表明,该图像检索方法有较高的准确性,并且有不错的检索效果。

Abstract: Active learning has been proved to be a key technique for improving Content-Based Image Retrieval(CBIR) performance.Relevance feedback technique can effectively reduce the cost of labeling.An active learning algorithm is put forward,weighted Co-Active SVM,to improve the performance of selective sampling in image retrieval.Color and texture are naturally considered as sufficient and uncorrelated views of an image;calculate the weight of color and texture feature space separately.SVM classifiers are learned in color and texture feature subspaces,respectively and the unlabeled data are classified.In order to reduce redundancy between these examples,K-means based active selection criterion is proposed to select images for user’s feedback.The experimental results show that the proposed algorithm has a higher accuracy,and has the better retrieval effect.

中图分类号: