The talk will be devoted to a new class of higher-order anisotropic total variation regularizers, with applications to image denoising and to digital elevation maps reconstruction. Our model can mix isotropic and anisotropic regularization at various derivative orders, which helps recovering the intrinsic structure of images and surfaces. I will describe a primal-dual approach for the numerical implementation of the model. The talk will be illustrated with various numerical examples.

This presentation is part of Minisymposium “MS26 - New trends in inpainting”
organized by: Yann Gousseau (Telecom ParisTech) , Simon Masnou (Université Lyon 1) .

Authors:

Simone Parisotto (University of Cambridge)

Simon Masnou (Université Lyon 1)

Carola-Bibiane Schönlieb (University of Cambridge)

Keywords:

image reconstruction, image representation, partial differential equation models