Disentangling Noise from Images: A Flow-Based Image Denoising Neural Network
The prevalent convolutional neural network (CNN)-based image denoising methods extract features of images to restore the clean ground truth, achieving high denoising accuracy. However, these methods may ignore the underlying distribution of clean images, inducing distortions or artifacts in denoisin...
Main Authors: | , , , , , |
---|---|
Format: | Article |
Language: | English |
Published: |
MDPI AG
2022-12-01
|
Series: | Sensors |
Subjects: | |
Online Access: | https://www.mdpi.com/1424-8220/22/24/9844 |
_version_ | 1797455301295734784 |
---|---|
author | Yang Liu Saeed Anwar Zhenyue Qin Pan Ji Sabrina Caldwell Tom Gedeon |
author_facet | Yang Liu Saeed Anwar Zhenyue Qin Pan Ji Sabrina Caldwell Tom Gedeon |
author_sort | Yang Liu |
collection | DOAJ |
description | The prevalent convolutional neural network (CNN)-based image denoising methods extract features of images to restore the clean ground truth, achieving high denoising accuracy. However, these methods may ignore the underlying distribution of clean images, inducing distortions or artifacts in denoising results. This paper proposes a new perspective to treat image denoising as a distribution learning and disentangling task. Since the noisy image distribution can be viewed as a joint distribution of clean images and noise, the denoised images can be obtained via manipulating the latent representations to the clean counterpart. This paper also provides a distribution-learning-based denoising framework. Following this framework, we present an invertible denoising network, FDN, without any assumptions on either clean or noise distributions, as well as a distribution disentanglement method. FDN learns the distribution of noisy images, which is different from the previous CNN-based discriminative mapping. Experimental results demonstrate FDN’s capacity to remove synthetic additive white Gaussian noise (AWGN) on both category-specific and remote sensing images. Furthermore, the performance of FDN surpasses that of previously published methods in real image denoising with fewer parameters and faster speed. |
first_indexed | 2024-03-09T15:52:27Z |
format | Article |
id | doaj.art-4b920101a5d04aa892375bcd0b0cd20b |
institution | Directory Open Access Journal |
issn | 1424-8220 |
language | English |
last_indexed | 2024-03-09T15:52:27Z |
publishDate | 2022-12-01 |
publisher | MDPI AG |
record_format | Article |
series | Sensors |
spelling | doaj.art-4b920101a5d04aa892375bcd0b0cd20b2023-11-24T17:56:16ZengMDPI AGSensors1424-82202022-12-012224984410.3390/s22249844Disentangling Noise from Images: A Flow-Based Image Denoising Neural NetworkYang Liu0Saeed Anwar1Zhenyue Qin2Pan Ji3Sabrina Caldwell4Tom Gedeon5The Research School of Computer Science, The Australian National University, Canberra, ACT 2600, AustraliaThe Research School of Computer Science, The Australian National University, Canberra, ACT 2600, AustraliaThe Research School of Computer Science, The Australian National University, Canberra, ACT 2600, AustraliaThe OPPO US Research, San Francisco, CA 94303, USAThe Research School of Computer Science, The Australian National University, Canberra, ACT 2600, AustraliaThe Research School of Computer Science, The Australian National University, Canberra, ACT 2600, AustraliaThe prevalent convolutional neural network (CNN)-based image denoising methods extract features of images to restore the clean ground truth, achieving high denoising accuracy. However, these methods may ignore the underlying distribution of clean images, inducing distortions or artifacts in denoising results. This paper proposes a new perspective to treat image denoising as a distribution learning and disentangling task. Since the noisy image distribution can be viewed as a joint distribution of clean images and noise, the denoised images can be obtained via manipulating the latent representations to the clean counterpart. This paper also provides a distribution-learning-based denoising framework. Following this framework, we present an invertible denoising network, FDN, without any assumptions on either clean or noise distributions, as well as a distribution disentanglement method. FDN learns the distribution of noisy images, which is different from the previous CNN-based discriminative mapping. Experimental results demonstrate FDN’s capacity to remove synthetic additive white Gaussian noise (AWGN) on both category-specific and remote sensing images. Furthermore, the performance of FDN surpasses that of previously published methods in real image denoising with fewer parameters and faster speed.https://www.mdpi.com/1424-8220/22/24/9844image denoisinginvertible networknormalizing flow |
spellingShingle | Yang Liu Saeed Anwar Zhenyue Qin Pan Ji Sabrina Caldwell Tom Gedeon Disentangling Noise from Images: A Flow-Based Image Denoising Neural Network Sensors image denoising invertible network normalizing flow |
title | Disentangling Noise from Images: A Flow-Based Image Denoising Neural Network |
title_full | Disentangling Noise from Images: A Flow-Based Image Denoising Neural Network |
title_fullStr | Disentangling Noise from Images: A Flow-Based Image Denoising Neural Network |
title_full_unstemmed | Disentangling Noise from Images: A Flow-Based Image Denoising Neural Network |
title_short | Disentangling Noise from Images: A Flow-Based Image Denoising Neural Network |
title_sort | disentangling noise from images a flow based image denoising neural network |
topic | image denoising invertible network normalizing flow |
url | https://www.mdpi.com/1424-8220/22/24/9844 |
work_keys_str_mv | AT yangliu disentanglingnoisefromimagesaflowbasedimagedenoisingneuralnetwork AT saeedanwar disentanglingnoisefromimagesaflowbasedimagedenoisingneuralnetwork AT zhenyueqin disentanglingnoisefromimagesaflowbasedimagedenoisingneuralnetwork AT panji disentanglingnoisefromimagesaflowbasedimagedenoisingneuralnetwork AT sabrinacaldwell disentanglingnoisefromimagesaflowbasedimagedenoisingneuralnetwork AT tomgedeon disentanglingnoisefromimagesaflowbasedimagedenoisingneuralnetwork |