Augmented Lagrangian method for total variation restoration with non-quadratic fidelity (Q534506)

The authors extend the augmented Lagrangian method for \(\text{TV-}L^2\) restoration. They first give the algorithms for the TV restoration with a general fidelity term and then apply these algorithms to recover blurry images corrupted by impulsive noise or Poisson noise. They show that the augmented Lagrangian method is extremely efficient for these two special cases. Besides, they provide the convergence analysis of these algorithms. In their method, they separately penalize the constraints and allow different penalty parameters for different constraints. Their method is different from those in the references.