Skip to main navigation Skip to search Skip to main content

Image classification by combining multiple SVMs

  • De Yuan Zhang*
  • , Bing Quan Liu
  • , Xiao Long Wang
  • , Li Juan Wang
  • *Corresponding author for this work
  • School of Computer Science and Technology, Harbin Institute of Technology
  • Hebei University

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

Abstract

In this paper, a novel framework is proposed for classifying images, which integrates several sets of Support Vector Machines(SVM) on multiple low level image features. In the proposed framework several global image features are extracted from the input images, and SVM using linear kernel with probability outputs are constructed on each feature. The outputs of the SVM classifiers are then combined by g2-fuzzy integral. The density value of the fuzzy integral for each classifier is trained by using grid searching algorithm. Compared with some current systems, our proposed framework demonstrates a promising performance for an image database of general-purpose images from Corel image library.

Original languageEnglish
Title of host publicationProceedings of the 7th International Conference on Machine Learning and Cybernetics, ICMLC
PublisherIEEE Computer Society
Pages68-73
Number of pages6
ISBN (Print)9781424420964
DOIs
StatePublished - 2008
Externally publishedYes
Event7th International Conference on Machine Learning and Cybernetics, ICMLC 2008 - Kunming, China
Duration: 12 Jul 200815 Jul 2008

Publication series

NameProceedings of the 7th International Conference on Machine Learning and Cybernetics, ICMLC
Volume1

Conference

Conference7th International Conference on Machine Learning and Cybernetics, ICMLC 2008
Country/TerritoryChina
CityKunming
Period12/07/0815/07/08

Keywords

  • Fuzzy integral
  • Global image feature
  • Image classification
  • Support Vector Machines

Fingerprint

Dive into the research topics of 'Image classification by combining multiple SVMs'. Together they form a unique fingerprint.

Cite this