Solution of Linear Systems Arising in Nonlinear Image Deblurring

摘要

This paper deals with the solution of large linear systems which arise when certain optimization methods are applied in image deblurring. An unconstrained penalized least squares minimization problem with total variation penalty is considered. Also addressed are several penalized least squares problems with nonnegativity constraints imposed on the solution. In each case, quasi-Newton techniques yield large structured linear systems. We investigate various preconditioning strategies to efficiently solve these linear systems.