Variational Networks: An Optimal Control Approach to Early Stopping Variational Methods for Image Restoration

Research output: Contribution to journalArticleResearchpeer-review

Abstract

We investigate a well-known phenomenon of variational approaches in image processing, where typically the best image quality is achieved when the gradient flow process is stopped before converging to a stationary point. This paradox originates from a tradeoff between optimization and modelling errors of the underlying variational model and holds true even if deep learning methods are used to learn highly expressive regularizers from data. In this paper, we take advantage of this paradox and introduce an optimal stopping time into the gradient flow process, which in turn is learned from data by means of an optimal control approach. As a result, we obtain highly efficient numerical schemes that achieve competitive results for image denoising and image deblurring. A nonlinear spectral analysis of the gradient of the learned regularizer gives enlightening insights about the different regularization properties.
Original languageEnglish
Pages (from-to)396-416
JournalJournal of Mathematical Imaging and Vision
Volume62
DOIs
Publication statusPublished - 11 Mar 2020

Fingerprint Dive into the research topics of 'Variational Networks: An Optimal Control Approach to Early Stopping Variational Methods for Image Restoration'. Together they form a unique fingerprint.

  • Cite this