Provably Convergent Plug & Play Linearized ADMM, applied to Deblurring Spatially Varying Kernels
This work addresses a computational bottleneck in inverse problem solving for image processing, offering a provably convergent method that is incremental in improving existing Plug & Play approaches.
The paper tackles the challenge of solving inverse problems like deblurring with spatially varying kernels by proposing a Plug & Play framework based on linearized ADMM that bypasses intractable proximal operators, demonstrating convergence and providing results on restoration tasks such as super-resolution and deblurring with non-uniform blur.
Plug & Play methods combine proximal algorithms with denoiser priors to solve inverse problems. These methods rely on the computability of the proximal operator of the data fidelity term. In this paper, we propose a Plug & Play framework based on linearized ADMM that allows us to bypass the computation of intractable proximal operators. We demonstrate the convergence of the algorithm and provide results on restoration tasks such as super-resolution and deblurring with non-uniform blur.