EM procedures using mean field-like approximations for Markov model-based image segmentation.
Gilles CeleuxFlorence ForbesNathalie PeyrardPublished in: Pattern Recognit. (2003)
Keyphrases
- em algorithm
- closed form
- expectation maximization
- free energy
- markov chain
- variational methods
- variational approximation
- maximum likelihood
- partition function
- markov model
- belief networks
- expectation maximisation
- posterior distribution
- bayesian inference
- generative model
- markov random field
- gaussian mixture model
- conditional independence
- mixture model
- unsupervised learning
- linear complexity
- statistical mechanics
- semi markov
- coordinate ascent
- image segmentation
- computationally tractable
- loopy belief propagation
- approximation methods
- markov process
- upper bound