Variational modeling is a powerful framework to address inverse imaging problems that enables to incorporate prior knowledge and to make use of efficient optimization tools, while offering theoretical guaranties. Regularized models based on gradient penalizations are very successful in image processing, such as the popular total variation formulations that have been proposed in the last decades. This symposium aims at giving a representative sample of such recent developments for applications to image and point-cloud restoration and segmentation, with new definitions of the total variation relying on original discrete formulations and adaptive non-local schemes, or combined with local regularity estimation.

Symmetric upwind scheme for discrete Non-Local Total Variation. Applications in image and point-cloud processing.

Sonia Tabti (Université de Caen, CNRS)

Regularized non-local Total Variation and application in image restoration

Zhi Li (Department of Computational Mathematics, Science and Engineering (CMSE) Michigan State University )

Combining Local Regularity Estimation and Total Variation Optimization for Scale-Free Texture Segmentation

Pascal Barbara (Laboratoire de Physique de l'ENS de Lyon)

Semi-Linearized Proximal Alternating Minimization for a Discrete Mumford-Shah Model

Marion Foare (Laboratoire de Physique, ENS Lyon)

Organizers:

Rabin Julien (CNRS, Normandie Univ.)

Sonia Tabti (Université de Caen, CNRS)

Keywords:

image denoising, image segmentation, variational models