An augmented Lagrangian algorithm for total bounded variation regularization based image deblurring

The augmented Lagrangian strategy has recently emerged as an important methodology for image processing problems. In this paper, based on this strategy, we propose a new projected gradient algorithm for image deblurring with total bounded variation regularization. The convergence property of our algorithm is discussed. Numerical experiments show that the proposed algorithm can yield better visual quality than the Rudin-Osher-Fatemi (ROF) method and the split Bregman iteration method.