Skip to main navigation Skip to search Skip to main content

Saliency-based region log covariance feature for image copy detection

  • Xin He
  • , Huiyun Jing
  • , Qi Han
  • , Xiamu Niu
  • Harbin Institute of Technology

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

This paper introduces a novel global feature-based image copy detection approach. Firstly, Space-based and Object-based saliency detection methods are combined to generate salient region represented by an ellipse. Then the covariance matrix of various image features extracted from the elliptically salient region is formed and log covariance matrix is applied on the covariance matrix for low computational complexity. 28 independent numbers from log covariance matrix are regarded as the region feature vector, the similarity of which can be measured by L2 norm. The experimental results show that our proposed approach achieves similar or better performance than GIST and log covariance matrix based SCOV for image copy detection.

Original languageEnglish
Title of host publicationDigital Forensics and Watermaking - 11th International Workshop, IWDW 2012, Revised Selected Papers
Pages327-335
Number of pages9
DOIs
StatePublished - 2013
Event11th International Workshop on Digital Forensics and Watermaking, IWDW 2012 - Shanghai, China
Duration: 31 Oct 20123 Nov 2012

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume7809 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

Conference11th International Workshop on Digital Forensics and Watermaking, IWDW 2012
Country/TerritoryChina
CityShanghai
Period31/10/123/11/12

Keywords

  • Global feature
  • Image copy detection
  • Log covariance matrix
  • Saliency detection

Fingerprint

Dive into the research topics of 'Saliency-based region log covariance feature for image copy detection'. Together they form a unique fingerprint.

Cite this