Mat-1.3656 Seminar on
numerical analysis and computational science
Monday, Nov 18, 2013, room M233 at 14.15, Eirola & Stenberg
Lauri Harhanen, Aalto!
Edge-preserving priorconditioning for linear ill-posed problems:
We present a method for solving large-scale linear inverse problems
regularized with a nonlinear, edge-preserving penalty term such as the
total variation or Perona–Malik. In the proposed scheme, the
nonlinearity in the regularizer is handled with lagged diffusivity
fixed point iteration which involves solving a large-scale linear least
squares problem in each iteration. Because of the notoriously slow
convergence of Krylov methods for problems of this type, we propose to
accelerate it by means of priorconditioning. Priorconditioning is a
technique which embeds the information contained in the prior or
regularizer directly into the forward operator through preconditioning
and leads to tremendously accelerated convergence.