Iterative methods for total variation denoising. (English) Zbl 0847.65083

The paper is concerned with computing the minimization of the total variation (TV)-penalized least squares functional. A fixed point algorithm is presented and compared with other minimization schemes. This is an alternative approach to minimizing the functional considered in the paper, called “lagged diffusivity fixed point iteration” and denoted by FP.
A variant of the cell-centered finite difference multigrid method of R. E. Ewing and J. Shen [A multigrid algorithm for the cell-centered finite difference scheme. Proc. 6th Copper Mountain Conf. Multigrid Methods, April 1993, NASA Conf. Publ. 3224 (1993)] is implemented for solving the (large and sparse) linear subproblems. In the last section, numerical results are performed. A numerical comparison of three methods applied to minimize the TV-penalized least squares functional is presented: the FP iteration, Newton’s method, and the steepest descent method. The results obtained by the three methods are close.


65N55 Multigrid methods; domain decomposition for boundary value problems involving PDEs
65H10 Numerical computation of solutions to systems of equations
65N06 Finite difference methods for boundary value problems involving PDEs


Full Text: DOI Link