Fast Unsupervised Tensor Restoration via Low-rank Deconvolution

TL;DR

This paper introduces an extension of Low-rank Deconvolution with differential regularization, enabling efficient signal restoration that rivals deep learning methods in performance and computational cost.

Contribution

It proposes a novel extension of Low-rank Deconvolution with differential regularization, incorporating TV and integral priors for improved signal restoration.

Findings

Competitive performance in image denoising and video enhancement

Low computational cost compared to deep learning methods

Effective integration of TV and integral priors

Abstract

Low-rank Deconvolution (LRD) has appeared as a new multi-dimensional representation model that enjoys important efficiency and flexibility properties. In this work we ask ourselves if this analytical model can compete against Deep Learning (DL) frameworks like Deep Image Prior (DIP) or Blind-Spot Networks (BSN) and other classical methods in the task of signal restoration. More specifically, we propose to extend LRD with differential regularization. This approach allows us to easily incorporate Total Variation (TV) and integral priors to the formulation leading to considerable performance tested on signal restoration tasks such image denoising and video enhancement, and at the same time benefiting from its small computational cost.