A Metric Framework for Texture Measurements

Fredrik Georgsson, Patrik Eklund
Department of Computing Science, Umeå University
901 87 Umeå
E-mail: {fredrikg,peklund}@cs.umu.se

Abstract:

A metric framework for measuring the distances between two textures is proposed. The metrics operate on subsets of images, but by using a transformation, , and by constructing two granular metrics operating on Y and F(Y,M), flexible and accurate metrics are obtained. For some textures, the metrics are better than a human observer in distinguishing which textures are similiar and which are not.

• Key words
• Introduction
• Metric image spaces
  • Granulation using range space
  • Granulation using domain space
• The transformation
• Results
• Conclusion
• Future work
• References
• About this document ...