Novel method to learn region weighting from relevance feedback in image retrieval

Yong Ge, Richang Hong, Xiuqing Wu, Xinmei Tian

Research output: Contribution to journalArticlepeer-review


In region-based image retrieval, relevance feedback and region weighing are two crucial problems. The former can improve the accuracy of all CBIR systems, if properly employed. The latter is an important factor in computing region-based similarity of two images. This paper proposes a novel region weighing method based on the users' feedback information which nicely combines both the positive example (PE) and the negative example (NE). The use of NE is to complement the use of PE, thereby improving retrieval accuracy. Through region importance (RI) calculation, the method can directly discard some less important regions for the RIs are zeros. The test results show that this method is effective in region-based image retrieval.

Original languageEnglish (US)
Pages (from-to)1643-1648
Number of pages6
JournalJournal of Computational Information Systems
Issue number4
StatePublished - Apr 2007
Externally publishedYes


  • CBIR
  • NE
  • PE
  • Region Importance
  • Relevance Feedback

ASJC Scopus subject areas

  • Information Systems
  • Computer Science Applications


Dive into the research topics of 'Novel method to learn region weighting from relevance feedback in image retrieval'. Together they form a unique fingerprint.

Cite this