We consider the Blind Source Separation (BSS) of images whose prior distributions are modelled through a compound Gauss-Markov modeling with either a hidden contour process, or a hidden classification labels modelled with a Potts distribution.The joint estimation of sources, hidden variables, mixing matrix and all the hyperparameters of the model (noise covariance, means and variances of the pixels in each region) can be done through the Gibbs sampling of the joint posterior probability law of these unknowns, given the observed mixed data. In previous works, we had implemented this general algorithm. However the huge complexity and cost of this posterior law limit its use in practical applications. In this paper, we propose a Mean Field Approximation (MFA) approach which consists in approximating this joint posterior by a separable one whose sampling will be much easier and less expensive. We then compare the relative performances of the models with regards of the MFA.
[1]
Jun Zhang,et al.
The mean field theory in EM procedures for blind Markov random field image restoration
,
1993,
IEEE Trans. Image Process..
[2]
Guillaume Rellier.
Analyse de textures dans l'espace hyperspectral par des méthodes probabilistes. (Texture analysis in the hyperspectral space by probabilistic methods)
,
2002
.
[3]
E. Salerno,et al.
BLIND SEPARATION OF AUTO-CORRELATED IMAGES FROM NOISY MIXTURES USING MRF MODELS
,
2003
.
[4]
Florence Forbes,et al.
Hidden Markov Random Field Model Selection Criteria Based on Mean Field-Like Approximations
,
2003,
IEEE Trans. Pattern Anal. Mach. Intell..
[5]
Hichem Snoussi,et al.
Fast joint separation and segmentation of mixed images
,
2004,
J. Electronic Imaging.