Gravitational weighted fuzzy c-means with application on multispectral image segmentation

Ahmed Ben Said, Rachid Hadjidj, Sebti Foufou

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

This paper presents a novel clustering approach based on the classic Fuzzy c-means algorithm. The approach is inspired from the concept of interaction between objects in physics. Each data point is regarded as a particle. A specific weight is associated with each data particle depending on its interaction with other particles. This interaction is induced by attraction forces between pairs of particles and the escape velocity from other particles. Classification experiments using two data sets from UCI repository demonstrate the outperformance of the proposed approach over other clustering algorithms. In addition, results demonstrate the effectiveness of the proposed scheme for segmentation of multispectral face images.

Original languageEnglish (US)
Title of host publication2014 4th International Conference on Image Processing Theory, Tools and Applications, IPTA 2014
EditorsKhalifa Djemal
PublisherInstitute of Electrical and Electronics Engineers Inc.
ISBN (Electronic)9781479964611
DOIs
StatePublished - Jan 6 2015
Event4th International Conference on Image Processing Theory, Tools and Applications, IPTA 2014 - Paris, France
Duration: Oct 14 2014Oct 17 2014

Publication series

Name2014 4th International Conference on Image Processing Theory, Tools and Applications, IPTA 2014

Other

Other4th International Conference on Image Processing Theory, Tools and Applications, IPTA 2014
CountryFrance
CityParis
Period10/14/1410/17/14

Keywords

  • Clustering
  • Gravity theories
  • Multispectral images
  • Segmentation

ASJC Scopus subject areas

  • Computer Vision and Pattern Recognition
  • Computer Science Applications

Fingerprint Dive into the research topics of 'Gravitational weighted fuzzy c-means with application on multispectral image segmentation'. Together they form a unique fingerprint.

Cite this