In texture segmentation, features must be firstly extracted in the mixture-of-Gaussian (MOG) models. In this paper, we combine MOG model with Gauss Markov random field (GMRF) model and get a unification model. This unified model takes interaction coefficients of neighbor pixels as parameters. We derivate a set of parameters estimation equations by expectation-maximization (EM) algorithms and apply them to a two-class texture segmentation problem. Experimental results show the efficiencies and strengths of the model.
A unified model of GMRF and MOG for image segmentation
IEEE International Conference on Image Processing 2005 ; 3 ; III-1140
2005-01-01
257434 byte
Conference paper
Electronic Resource
English
A Unified Model of GMRF and MOG for Image Segmentation
British Library Conference Proceedings | 2005
|Unsupervised urban area extraction from SAR imagery using GMRF
British Library Online Contents | 2006
|Unsupervised urban area extraction from SAR imagery using GMRF
British Library Online Contents | 2008
|Mining Very High Resolution InSAR Data Based on Complex-GMRF Cues and Relevance Feedback
British Library Conference Proceedings | 2012
|A Unified Framework for Segmentation-Assisted Image Registration
British Library Conference Proceedings | 2006
|