Definition of a Spatial Entropy and its Use for Texture Discrimination

This paper presents a new definition of a spatial entropy mainly based on the Markov Random Field (MRF) properties. Starting with the study of the entropy proposed in [1] for the Potts model, we establish a specific property of the entropy in this special case, and derive the analytical expressions for a 4-connexity neighborhood. Inspired by the previous property, we propose a new definition mainly based on an heterogeneity measure of the neighborhood. This definition is then used to analyze a SAR (Synthetic Aperture Radar) image and it is shown to be able to discriminate different types of textures.

• Type: Conference
• Year: 2000
• Where: ICIP
• Authors: Florence Tupin, Henri Maître, Marc Sigelle
• Keywords: Heterogeneity Measure, ICIP 2000, Image Processing, Markov Random Field, Synthetic Aperture Radar