A matrix-free method for regularisation with unrestricted variables
Abstract
In this thesis a method for the partially norm constrained least squares problem is presented. The method relies on a large-scale trust-region solver and has a low storage requirement. A combination of image misalignment and the inverse problem deblurring illustrates the use of the method.
Publisher
The University of BergenCopyright
The authorCopyright the author. All rights reserved