1: \begin{abstract}

2:   In this work, we consider a class of differentiable criteria for sparse

3:   image computing problems, where a nonconvex regularization is applied to an

4:   arbitrary linear transform of the target image. As special cases, it includes

5:   edge preserving measures or frame-analysis potentials commonly used in image

6:   processing. As shown by our asymptotic results, the $\ell_2-\ell_0$ penalties

7:   we consider may be employed to provide approximate solutions to

8:   $\ell_0$-penalized optimization problems. One of the advantages of the

9:   proposed approach is that it allows us to derive an efficient

10:   Majorize-Minimize subspace algorithm. The convergence of the algorithm is

11:   investigated by using recent results in nonconvex optimization.  The fast

12:   convergence properties of the proposed optimization method are illustrated

13:   through image processing examples. In particular, its effectiveness is

14:   demonstrated on several data recovery problems.

15: \end{abstract}

16: