Published on Sat Oct 03 2020

Async-RED: A Provably Convergent Asynchronous Block Parallel Stochastic Method using Deep Denoising Priors

Yu Sun, Jiaming Liu, Yiran Sun, Brendt Wohlberg, Ulugbek S. Kamilov

Red is a framework for solving inverse problems by integrating advanced denoisers as image priors. Current RED algorithms are inadequate for parallel processing on multicore systems. ASYNC-RED is significantly faster than its serial counterparts for large-scale inverse problems.

0
0
0
Abstract

Regularization by denoising (RED) is a recently developed framework for solving inverse problems by integrating advanced denoisers as image priors. Recent work has shown its state-of-the-art performance when combined with pre-trained deep denoisers. However, current RED algorithms are inadequate for parallel processing on multicore systems. We address this issue by proposing a new asynchronous RED (ASYNC-RED) algorithm that enables asynchronous parallel processing of data, making it significantly faster than its serial counterparts for large-scale inverse problems. The computational complexity of ASYNC-RED is further reduced by using a random subset of measurements at every iteration. We present complete theoretical analysis of the algorithm by establishing its convergence under explicit assumptions on the data-fidelity and the denoiser. We validate ASYNC-RED on image recovery using pre-trained deep denoisers as priors.

Fri Jun 05 2020
Machine Learning
Scalable Plug-and-Play ADMM with Convergence Guarantees
plug-and-play priors (PnP) is a broadly applicable methodology for exploiting statistical priors specified as denoisers. Current PnP algorithms are impractical in large-scale settings due to their heavy computational and memory requirements. This work addresses this issue by proposing an incremental
1
0
1
Sat Jun 20 2020
Computer Vision
A Fast Stochastic Plug-and-Play ADMM for Imaging Inverse Problems
PnP stochastic gradient descent methods have recently been proposed and shown improved performance in some imaging applications. Current PnP methods need to frequently compute the image denoisers which can be computationally expensive.
0
0
0
Mon Jan 29 2018
Computer Vision
Learning-based Image Reconstruction via Parallel Proximal Algorithm
In the past decade, sparsity-driven regularization has led to advancement of image reconstruction algorithms. We propose to generalize TV regularization by replacing the l1-penalty with an alternative prior that is trainable. The proposed framework does not require additional inner iterations.
0
0
0
Tue Oct 22 2019
Computer Vision
The Practicality of Stochastic Optimization in Imaging Inverse Problems
Stochastic gradient descent algorithms have been shown in the machine learning literature to have optimal complexities in theory. Surprisingly, many of such methods fail to converge faster than the accelerated deterministic gradient methods. The proposed method has fast convergence rate in practice, and is able to handle non-smooth regularization terms.
0
0
0
Wed Sep 12 2018
Computer Vision
An Online Plug-and-Play Algorithm for Regularized Image Reconstruction
plug-and-play priors (PnP) is a powerful framework for regularizing imaging problems by using advanced denoisers. PnP algorithms achieve state-of-the-art performance in a range of imaging applications.
0
0
0
Thu May 05 2016
Computer Vision
Plug-and-Play ADMM for Image Restoration: Fixed Point Convergence and Applications
Alternating direction method of multiplier (ADMM) is a widely used algorithm for image restoration. One critical feature of the ADMM algorithm is its modular structure. It allows one to plug in any off-the-shelf image denoising algorithm for a subproblem. This type of ADMM algorithms is
0
0
0