Catalogue of Artificial Intelligence Techniques
Categories: Knowledge Representation
Author(s): Judea Pearl
A probabilistic description of a system of variables that interact locally, organised as vertices in a graph. A Markov field encodes the assumption that each variable is influenced by only its nearest neighbours. In image processing, Markov fields are used to represent the correlations between adjacent picture cells. In Neural Networks, Markov fields represent the energy associated with each state.
- Geman, S. and Geman, D., Stochastic relaxation, Gibbs distributions and the Bayesian restoration of images
IEEE Trans on PAMI 6 (1984), 721--741.