Hidden fuzzy Markov chain model with K discrete classes

Ahmed Gamal Eldin (1, 2) , Fabien Salzenstein (2) , Christophe Collet (2)

1 ARIANA - Inverse problems in earth monitoring
2 LSIIT / MIV - Laboratoire des Sciences de l'Image, de l'Informatique et de la Télédétection, équipe MIV

This paper deals with a new unsupervised fuzzy Bayesian segmentation method based on the hidden Markov chain model, in order to separate continuous from discrete components in the hidden data. We present a new F-HMC (fuzzy hidden Markov chain) related to three hard classes, based on a general extension of the previously algorithms proposed. For a given observation, the hidden variable owns a density according to a measure containing Dirac and Lebesgue components. We have performed our approach in the multispectral context. The hyper-parameters are estimated using a Stochastic Expectation Maximization (SEM) algorithm. We present synthetic simulations and also segmentation results related to real multi-band data.

Mots clés: fuzzy segmentation; hidden Markov chain

Domaines: Traitement du signal et de l'image [eess.SP]; Traitement du signal et de l'image [eess.SP]

Ahmed Gamal Eldin, Fabien Salzenstein, Christophe Collet. Hidden fuzzy Markov chain model with K discrete classes. Information Sciences Signal Processing and their Applications (ISSPA), May 2010, Kuala Lumpur, Malaysia. ⟨hal-00616372⟩