Accuate numerical method for total variation-based image deblurring

Total variation regularization functional is often used to restore an image from blurred image. TV-based method is a nonlinear image reconstruction method. The schemes to handle TV nonlinearity require the solution of a system of linear equations. The systems which arise in TV-based deblurring present some very serious numerical challenges. In this paper, we present a new solver which is block SSOR-based preconditioned conjugate gradient scheme. It allows for the fast solution of these very large linear systems. Our experiments indicate that our method is capable of producing highly accurate solutions and preserving more edge and object boundaries.