Deconvolution-segmentation for textured images

Abstract

The paper tackles the problem of joint deconvolution and segmentation specifically for textured images. The images are composed of patches of textures that belong to a set of K possible classes. Each class of image is described by a Gaussian random field and the classes are modelled by a Potts field. The method relies on a hierarchical model and a Bayesian strategy to jointly estimate the labels, the textured images as well as the hyperparameters. An important point is that the parameter of the Potts field is also estimated. The estimators are designed in an optimal manner (marginal posterior maximizer for the labels and posterior mean for the other unknowns). They are computed based on a convergent procedure, from samples of the posterior obtained through an MCMC algorithm (Gibbs sampler including PerturbationOptimization). A first numerical evaluation provides encouraging results despite the strong difficulty of the problem.

Unsupervised joint deconvolution and segmentation method for textured images: a Bayesian approach and an advanced sampling algorithm

“…The procedure is identical to the one presented in our previous papers [40,41,57]. The reader is invited to consult [29,[42][43][44][45][46] for alternatives and complementary results.…”