Applying the Information Bottleneck Principle to Unsupervised Clustering of Discrete and Continuous Image Representations

  • Authors:
  • Shiri Gordon;Hayit Greenspan;Jacob Goldberger

  • Affiliations:
  • -;-;-

  • Venue:
  • ICCV '03 Proceedings of the Ninth IEEE International Conference on Computer Vision - Volume 2
  • Year:
  • 2003

Quantified Score

Hi-index 0.00

Visualization

Abstract

In this paper we present a method for unsupervised clustering ofimage databases. The method is based on a recently introducedinformation-theoretic principle, the information bottleneck (IB)principle. Image archives are clustered such that the mutualinformation between the clusters and the image content is maximallypreserved. The IB principle is applied to both discrete andcontinuous image representations, using discrete image histogramsand probabilistic continuous image modeling based on mixture ofGaussian densities, respectively. Experimental results demonstratethe performance of the proposed method forimage clustering on alarge image database. Several clustering algorithms derived fromthe IB principle are explored and compared.