Computational Analysis Seminar

Sergio Amat Plata

(Universidad Politécnica de Cartagena)

"Variational image denoising using PDE and multiresolution frameworks"


Date: Tue, May 4, 2010
Time: 15:45
Place: Research I Seminar Room

Abstract: The Total Variation minimization model has been used in many applications related to image processing. It was introduced as a PDE-based algorithm for edge-preserving noise removal by Rudin, Osher and Fatemi. The associated Euler-Lagrange equation includes a highly nonlinear term. The main difficulty in the analysis and numerical approximation of these equations is the linearization of this term. In this paper some approximation procedures are studied. The motivation of our approaches are based on two aspects: first, to obtain simple models, and, secondly, retain the main feature of the TV-model only in regions near discontinuities. We study both PDE and multiresolution frameworks.