2009 | OriginalPaper | Buchkapitel
Variational Methods for Denoising Matrix Fields
verfasst von : S. Setzer, G. Steidl, B. Popilka, B. Burgeth
Erschienen in: Visualization and Processing of Tensor Fields
Verlag: Springer Berlin Heidelberg
Aktivieren Sie unsere intelligente Suche, um passende Fachinhalte oder Patente zu finden.
Wählen Sie Textabschnitte aus um mit Künstlicher Intelligenz passenden Patente zu finden. powered by
Markieren Sie Textabschnitte, um KI-gestützt weitere passende Inhalte zu finden. powered by
The restoration of scalar-valued images via minimization of an energy functional is a well-established technique in image processing. Recently, higher-order methods have proved their advantages in edge preserving image denoising. In this chapter, we transfer successful techniques like the minimization of the Rudin-Osher-Fatemi functional and the infimal convolution to matrix fields, where our functionals couple with different matrix channels. For the numerical computation, we use second-order cone programming. Moreover, taking the operator structure of matrices into account, we consider a new operator-based regularization term. This is the first variational approach for denoising tensor-valued data that takes the operator structure of matrices, in particular the operation of matrix multiplication into account. Using matrix differential calculus, we deduce the corresponding Euler-Lagrange equation and apply it for the numerical solution by a steepest descent method.