期刊文献+

基于改进的排序学习的图片检索算法研究 被引量:1

Learning to Rank Based Approach for Image Searching
下载PDF
导出
摘要 图片检索是图片共享社会网络中的重要研究内容之一。传统的图片检索方法往往通过对用户输入的关键字和图片的文本描述加以匹配来进行图片检索。由于文本信息存在歧义性,图片的文本描述十分困难,因此检索结果的准确性低。为了提高图片检索的准确性,提出了基于排序学习的图片检索方法。将每幅图片通过多种特征描述符进行描述,当用户的输入为图片时,通过对比查询图片和图片库中图片的相似性进行图片检索。采用支持向量机和关联规则两种学习方法对特征描述符的权重组合进行学习,并提出了相应的学习算法。实验表明,提出的基于学习的图片检索方法与相关图片检索方法相比具有更高的准确性。此外,应用支持向量机和关联规则两种方法对分类函数进行学习时,由于两种算法通过相同的数据实例对图片描述符的权重进行学习,因此得到的结果是相关的。 Image searching is one of the most important researches in image sharing based social networks. Traditional image searching methods usually compare the user keywords and the textual description of images in database while searching. Because the textual description is ambiguous, the abstracting of text for images is very hard, and thus the accuracy of image searching is low. In order to improve the accuracy of image searching, this paper proposed a learning to rank based approach. We described each image as a combination of multiple feature descriptors, and compared the similarity of the query and the image in database while users input a query of image. We applied association rules and sup- port vector machine to learn the weight of each feature descriptor, and proposed corresponding learning algorithms. The experiments show that the proposed image searching approach is more accurate than related works while retrieving image for users. In addition,while using association rule and support vector machine to learn the classification functions, the two algorithms use the same instances to learn the weight of each feature descriptor, so they are relevant.
出处 《计算机科学》 CSCD 北大核心 2015年第12期275-277,306,共4页 Computer Science
基金 江西省博士研究生创新项目科研基金(YC2011-B026)资助
关键词 图片检索 排序学习 支持向量机 关联规则 Image searching, Learning to rank, Support vector machine, Association rule
  • 相关文献

参考文献3

二级参考文献31

  • 1吴洪,卢汉清,马颂德.基于内容图像检索中相关反馈技术的回顾[J].计算机学报,2005,28(12):1969-1979. 被引量:52
  • 2Golder S A, Huberman B A. Usage patterns of collaborative tagging systems [ J ]. Journal of Information Science, 2006,32(2): 198-208.
  • 3Schmitz P. Inducing ontology from flickr tags[ C ]//Proceedings of the Workshop on Collaborative Web Tagging at WWW. Edinburgh, UK: ACM Press,2006.
  • 4Schmitz C, Hotho A, Jaschke R, et al. Mining association rules in folksonomies[ C]// Proceedings of the 10th IFCS Conference. Berlin, Heidelberg, German: Springer, 2006, 260-270.
  • 5Mika P. Ontologies are us : A unified model of social networks and semantics[ J]. Journal of Web Semantics, 2007, 5 (1): 5-15.
  • 6Leydesdorff L, Vaughan L. Co-occurrence matrices and their applications in information science: extending ACA to the web environment[ J].Journal of the American Society for Information Science and Technology, 2006, 57(12): 1616-1628.
  • 7Jing F, Wang C, Yao Y, et al. Igroup: Web image search results clustering [ C ]//Proceedings of the 14th Annual ACM International Conference on Multimedia. New York, USA: ACM, press, 2006, 377-384.
  • 8Pothen A, Simon H D, Liou K P. Partitioning sparse matrices with eigenvectnrs of graphs [ J ]. SIAM Journal on Matrix Analysis and Applications, 1990, 11 (3) : 430-452.
  • 9White S, Smyth P. A spectral clustering approach to finding communities in graphs [ C ] //Proceedings of the Fifth SIAM International Conference on Data Mining. Newport Beach, CA, USA: SIAM, 2005, 274-285.
  • 10Miller G A, Beckwith R, Fellbaum C, et al. Introduction to wordnet : an on- line lexical database [ J ]. International Journal of Lexicography, 1990, 3(4) :235-244.

共引文献20

同被引文献10

引证文献1

二级引证文献5

相关作者

内容加载中请稍等...

相关机构

内容加载中请稍等...

相关主题

内容加载中请稍等...

浏览历史

内容加载中请稍等...
;
使用帮助 返回顶部