计算机工程与应用 ›› 2007, Vol. 43 ›› Issue (36): 137-140.

• 网络、通信与安全 • 上一篇    下一篇

基于多特征融合和Adaboost算法的图像检索

朱守业   

  1. 徐州师范大学 信息传播学院,江苏 徐州 221009
  • 收稿日期:1900-01-01 修回日期:1900-01-01 出版日期:2007-12-21 发布日期:2007-12-21
  • 通讯作者: 朱守业

Image retrieval based on multi-feature integration and Adaboost relevance feedback

ZHU Shou-ye   

  1. School of Information and Media,Xuzhou Normal University,Xuzhou,Jiangsu 221009,China
  • Received:1900-01-01 Revised:1900-01-01 Online:2007-12-21 Published:2007-12-21
  • Contact: ZHU Shou-ye

摘要: 提出了一种多特征融合的图像检索方法。首先将图像进行分块,并提取分块主色,然后采用主色直方图作为图像的颜色特征。同时,提出采用Gabor小波描述图像的纹理特征,采用小波矩描述形状特征,最后将三种不同特征进行融合的检索方法。为了提高图像检索的准确度,提出Adaboost的相关反馈算法,在反馈过程中,Adaboost算法对特征进行降维,加快检索的速度。最后分别给出基于单一特征,特征融合和相关反馈方法的查准率和查全率,并对实验结果进行分析。

关键词: 图像检索, 特征融合, 相关反馈, Adaboost算法

Abstract: A image retrieval method based on multi-feature integration is proposed.Images are firstly partitioned into equal-sized sub-images,then the main color of each block is extracted and the main color histogram is computed as the image’s color feature.Gabor wavelet is proposed to describe the texture feature,and wavelet invariant moments is used to describe shape feature.In order to improve the precision of image retrieval,a relevance feedback mechanism,based on Adaboost method is invoked.Adaboost method is used to realize the dimensionality reduction and improve retrieval speed.Finally,the precisions and recalls of single feature,multi-feature integration and relevance feedback are calculated and compared.

Key words: image retrieval, feature integration, relevance feedback, Adaboost method