Quantitative Description of Spatially Homogeneous Textures by Characteristic Grey Level Co-Occurrences

Gibbs random eld model with multiple pairwise pixel interactions describes each type of spatially homogeneous image textures in terms of a pixel neighbourhood and Gibbs potentials that specify the geometric structure and quantitative strengths of interactions, respectively. Each image is represented by a set of grey level co-occurrence histograms which are sucient statistics of the model. The histograms permit us to analytically approximate Gibbs potentials and recover most characteristic neighbourhood from a given training sample so that the model provides a theoretical framework for texture description by intergal histogram-based features.

• Type: Journal
• Year: 2000
• Where: Australian Journal of Intelligent
• Authors: Georgy Gimel'farb
• Keywords: Gibbs Random Field, Grey Level Co-occurrence, Markov Random Fields (MRF), Texture