We apply basic statistical reasoning to signal reconstruction by machine
learning -- learning to map corrupted observations to clean signals -- with a
simple and powerful conclusion: it is possible to learn to restore images by
only looking at corrupted examples, at performance at and sometimes exceeding
training using clean data, without explicit image priors or likelihood models
of the corruption. In practice, we show that a single model learns photographic
noise removal, denoising synthetic Monte Carlo images, and reconstruction of
undersampled MRI scans -- all corrupted by different processes -- based on
noisy data only.
Description
Noise2Noise: Learning Image Restoration without Clean Data
%0 Generic
%1 lehtinen2018noise2noise
%A Lehtinen, Jaakko
%A Munkberg, Jacob
%A Hasselgren, Jon
%A Laine, Samuli
%A Karras, Tero
%A Aittala, Miika
%A Aila, Timo
%D 2018
%K cs.CV cs.LG
%T Noise2Noise: Learning Image Restoration without Clean Data
%U http://arxiv.org/abs/1803.04189
%X We apply basic statistical reasoning to signal reconstruction by machine
learning -- learning to map corrupted observations to clean signals -- with a
simple and powerful conclusion: it is possible to learn to restore images by
only looking at corrupted examples, at performance at and sometimes exceeding
training using clean data, without explicit image priors or likelihood models
of the corruption. In practice, we show that a single model learns photographic
noise removal, denoising synthetic Monte Carlo images, and reconstruction of
undersampled MRI scans -- all corrupted by different processes -- based on
noisy data only.
@misc{lehtinen2018noise2noise,
abstract = {We apply basic statistical reasoning to signal reconstruction by machine
learning -- learning to map corrupted observations to clean signals -- with a
simple and powerful conclusion: it is possible to learn to restore images by
only looking at corrupted examples, at performance at and sometimes exceeding
training using clean data, without explicit image priors or likelihood models
of the corruption. In practice, we show that a single model learns photographic
noise removal, denoising synthetic Monte Carlo images, and reconstruction of
undersampled MRI scans -- all corrupted by different processes -- based on
noisy data only.},
added-at = {2021-08-19T13:29:24.000+0200},
author = {Lehtinen, Jaakko and Munkberg, Jacob and Hasselgren, Jon and Laine, Samuli and Karras, Tero and Aittala, Miika and Aila, Timo},
biburl = {https://www.bibsonomy.org/bibtex/2b651f05c22da1fa84e35fb645f21fe15/aerover},
description = {Noise2Noise: Learning Image Restoration without Clean Data},
interhash = {3cae266f4bcc8af23fd46c355fc9a0cb},
intrahash = {b651f05c22da1fa84e35fb645f21fe15},
keywords = {cs.CV cs.LG},
note = {cite arxiv:1803.04189Comment: Added link to official implementation and updated MRI results to match it},
timestamp = {2021-08-19T13:29:24.000+0200},
title = {Noise2Noise: Learning Image Restoration without Clean Data},
url = {http://arxiv.org/abs/1803.04189},
year = 2018
}