Pairwise Markov model applied to unsupervised image separation

Selwa Rafi (1, 2, 3) , Marc Castella (1, 2, 3) , Wojciech Pieczynski (1, 2, 3)

1 TSP - CITI - Communications, Images et Traitement de l'Information
2 SAMOVAR - Services répartis, Architectures, MOdélisation, Validation, Administration des Réseaux
3 CNRS - Centre National de la Recherche Scientifique

The paper deals with blind separation and recovery of a noisy mixture of two binary signals on two sensors. Such a model can be applied in the context of recovery of scanned documents subject to show-through and bleed-through effects. The problem can be considered as a blind source separation one. Due to a complex noise and data structure, it is tackled from the more general approach of Bayesian restoration. The data is assumed to follow a Pairwise Markov Chain model: it generalizes Hidden Markov Chain models but it still allows one to calculate the a posteriori distributions of the data. The Expectation-Maximization (EM) and Iterative Conditional Estimation (ICE) methods are considered for parameter estimation, yielding an unsupervised processing. Finally, simulations show the interest of our approach on simulated and real data.

Mots clés: Blind source separation; Image separation; Show-through removal; Pairwise Markov chain

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

Selwa Rafi, Marc Castella, Wojciech Pieczynski. Pairwise Markov model applied to unsupervised image separation. SPPRA '11 : The Eighth IASTED International Conference on Signal Processing, Pattern Recognition, and Applications, Feb 2011, Innsbruck, Austria. ⟨10.2316/P.2011.721-044⟩. ⟨hal-00714717⟩