1: \begin{abstract}

2: An explicit algorithm for the minimization of an $\ell_1$

3: penalized least squares functional, with non-separable $\ell_1$

4: term, is proposed. Each step in the iterative algorithm

5: requires four matrix vector multiplications and a single simple

6: projection on a convex set (or equivalently thresholding).

7: Convergence is proven and a $1/N$ convergence rate is derived

8: for the functional. In the special case where the matrix in the

9: $\ell_1$ term is the identity (or orthogonal), the algorithm

10: reduces to the traditional iterative soft-thresholding

11: algorithm. In the special case where the matrix in the

12: quadratic term is the identity (or orthogonal), the algorithm

13: reduces to a gradient projection  algorithm for the dual

14: problem.

15: 

16: By replacing the projection with a simple proximity operator, other convex non-separable penalties than those based on an $\ell_1$-norm can be handled as well.

17: \end{abstract}

18: