Accelerated Bregman method for linearly constrained \(\ell _1-\ell _2\) minimization (Q368088): Difference between revisions

The Bregman method and a linearized Bregman method for solving total variation-based image restoration are reviewed. The authors propose an accelerated algorithm for the Bregman method to solve the linearly constrained \(\ell _1-\ell _2\) minimization. The proposed accelerated Bregman method is equivalent to the accelerated augmented Lagrangian method. The global convergence rate of the proposed algorithm is established. Some numerical test results are presented to show that the proposed method is faster than the original Bregman method and the accelerated linearized Bregman method.