I will discuss our recent work on patch-based models that are adapted to specific image classes, or even specific scenes. These models take the form of minimum mean squared error (MMSE) patch-based denoisers, using Gaussian mixture priors. We illustrate their use beyond image denoising, in more general inverse problems, such as inpainting, deblurring, and hyperspectral super-resolution, using the recently introduced plug-and-play approach.

This presentation is part of Minisymposium “MS52 - A Denoiser Can Do Much More Than Just... Denoising (2 parts)”
organized by: Yaniv Romano (Technion - Israel Institute of Technology) , Peyman Milanfar (Google Research) , Michael Elad (The Technion - Israel Institute of Technology) .

Authors:

Jose Bioucas-Dias (Universidade de Lisboa, Instituto Superior Técnico (IST), Instituto de Telecomunicações (IT))

Afonso M. Teodoro (Technical University of Lisbon)

Mário Figueiredo (Instituto de Telecomunicações and IST, University of Lisboa)

Keywords:

bayesian methods, image deblurring, image enhancement, image reconstruction, inverse problems