PASCAL - Pattern Analysis, Statistical Modelling and Computational Learning

Disagreement loop and path creation/annihilation algoritms for planar Markov fields with applications to image segmentation
T. Schreiber and M.N.M. van Lieshout
(2007) EURANDOM, Eindhoven.


We introduce a class of Gibbs--Markov random fields that can be understood as discrete versions of coloured polygonal fields built on regular tessellations. Under an appropriate choice of Hamiltonian polygonal we show these fields enjoy striking properties including consistency and solvability by means of dynamic representations. For general Hamiltonians we develop disagreement loop as well as path creation and annihilation dynamics yielding efficient simulation algorithms with non-local updates. Applications to foreground-background segmentation problems are discussed.

EPrint Type:Other
Project Keyword:Project Keyword UNSPECIFIED
Subjects:Machine Vision
Learning/Statistics & Optimisation
Theory & Algorithms
ID Code:3068
Deposited By:Marie-Colette van Lieshout
Deposited On:29 November 2007