Motivate the proposal, clarifying why the topic of the proposed Special Poster Session is timely and relevant. Briefly illustrate how the session is organized, possibly emphasizing different aspects of the proposed topic and how they are covered.
The basic problem behind image restoration, image denoising or image deconvolution consists of estimating an unknown image from noisy measurements via a so-called inverse problem. In order to solve this generally ill-posed problem, the classical approach is to optimize a cost function depending on data fidelity and regularization
terms. This cost function can be for instance obtained by using the principles of maximum likelihood theory, penalized least squares or Bayesian inference. Its optimization requires the use of sophisticated algorithms based optimization theory or on Markov chain Monte Carlo methods. The aim of this special session is to present some recent advances in this area, with a particular interest in confronting optimization and simulation methods for the image processing problems introduced before.
Several sessions in signal and image processing conferences (ICASSP, ICIP, IGARSS, EUSIPCO,…) are devoted every year to inverse problems. A tutorial on“Inverse Problems Regularized by Sparsity”was for instance presented in ICASSP 2013 by Martin Vetterli. However, new challenges related to inverse problems have appeared with the recent advances in sparse signal/image processing involving very large scale, non-smooth, and convex/non-convex optimization. The objective of this special session is to present recent contributions in this area with a specific interest in optimization and simulation methods for image processing. Note that a recent special issue on stochastic simulation and optimization in signal processing is currently scheduled (march 2016) for the IEEE Journal of Selected Topics in Signal Processing.