Fast Frame-Based Image Deconvolution Using Variable Splitting and Constrained Optimization

TL;DR

This paper introduces a fast, second-order optimization algorithm for frame-based image deconvolution that outperforms existing methods in speed by leveraging variable splitting and augmented Lagrangian techniques.

Contribution

It presents a novel algorithm combining variable splitting and augmented Lagrangian methods with second-order information for efficient image deconvolution.

Findings

Algorithm is significantly faster than previous methods.

Effective use of second-order Hessian information.

Demonstrated on benchmark image deblurring problems.

Abstract

We propose a new fast algorithm for solving one of the standard formulations of frame-based image deconvolution: an unconstrained optimization problem, involving an $\ell_2$ data-fidelity term and a non-smooth regularizer. Our approach is based on using variable splitting to obtain an equivalent constrained optimization formulation, which is then addressed with an augmented Lagrangian method. The resulting algorithm efficiently uses a regularized version of the Hessian of the data fidelity term, thus exploits second order information. Experiments on a set of image deblurring benchmark problems show that our algorithm is clearly faster than previous state-of-the-art methods.