Multi-Domain Multi-Scale Diffusion Model for Low-Light Image Enhancement

Shang, K; Shao, M; Wang, C; Cheng, Y; Wang, S

Multi-Domain Multi-Scale Diffusion Model for Low-Light Image Enhancement

Shang, K Shao, M Wang, C

Cheng, Y Wang, S

Permalink

Publisher:: Association for the Advancement of Artificial Intelligence (AAAI)
Publication Type:: Conference Proceeding
Citation:: Proceedings of the AAAI Conference on Artificial Intelligence, 2024, 38, (5), pp. 4722-4730
Issue Date:: 2024-03-25

Closed Access

	Filename	Description	Size
	Multi-Domain Multi-Scale.pdf	Published version	18.18 MB	Adobe PDF	View/Open

Copyright Clearance Process

Recently Added
In Progress
Closed Access

This item is closed access and not available.

Full metadata record

Field	Value	Language
dc.contributor.author	Shang, K
dc.contributor.author	Shao, M
dc.contributor.author	Wang, C https://orcid.org/0000-0003-1297-768X
dc.contributor.author	Cheng, Y
dc.contributor.author	Wang, S
dc.date.accessioned	2025-03-03T00:02:12Z
dc.date.available	2025-03-03T00:02:12Z
dc.date.issued	2024-03-25
dc.identifier.citation	Proceedings of the AAAI Conference on Artificial Intelligence, 2024, 38, (5), pp. 4722-4730
dc.identifier.issn	2159-5399
dc.identifier.issn	2374-3468
dc.identifier.uri	http://hdl.handle.net/10453/185480
dc.description.abstract	Diffusion models have achieved remarkable progress in lowlight image enhancement. However, there remain two practical limitations: (1) existing methods mainly focus on the spatial domain for the diffusion process, while neglecting the essential features in the frequency domain; (2) conventional patch-based sampling strategy inevitably leads to severe checkerboard artifacts due to the uneven overlapping. To address these limitations in one go, we propose a Multi-Domain Multi-Scale (MDMS) diffusion model for low-light image enhancement. In particular, we introduce a spatialfrequency fusion module to seamlessly integrates spatial and frequency information. By leveraging the Multi-Domain Learning (MDL) paradigm, our proposed model is endowed with the capability to adaptively facilitate noise distribution learning, thereby enhancing the quality of the generated images. Meanwhile, we propose a Multi-Scale Sampling (MSS) strategy that follows a divide-ensemble manner by merging the restored patches under different resolutions. Such a multi-scale learning paradigm explicitly derives patch information from different granularities, thus leading to smoother boundaries. Furthermore, we empirically adopt the Bright Channel Prior (BCP) which indicates natural statistical regularity as an additional restoration guidance. Experimental results on LOL and LOLv2 datasets demonstrate that our method achieves state-of-the-art performance for the low-light image enhancement task.
dc.language	en
dc.publisher	Association for the Advancement of Artificial Intelligence (AAAI)
dc.relation.ispartof	Proceedings of the AAAI Conference on Artificial Intelligence
dc.relation.isbasedon	10.1609/aaai.v38i5.28273
dc.rights	info:eu-repo/semantics/closedAccess
dc.title	Multi-Domain Multi-Scale Diffusion Model for Low-Light Image Enhancement
dc.type	Conference Proceeding
utslib.citation.volume	38
pubs.organisational-group	University of Technology Sydney
pubs.organisational-group	University of Technology Sydney/Faculty of Engineering and Information Technology
pubs.organisational-group	University of Technology Sydney/Faculty of Engineering and Information Technology/School of Computer Science
utslib.copyright.status	closed_access	*
dc.date.updated	2025-03-03T00:02:09Z
pubs.issue	5
pubs.publication-status	Published
pubs.volume	38
utslib.citation.issue	5

Abstract:

Diffusion models have achieved remarkable progress in lowlight image enhancement. However, there remain two practical limitations: (1) existing methods mainly focus on the spatial domain for the diffusion process, while neglecting the essential features in the frequency domain; (2) conventional patch-based sampling strategy inevitably leads to severe checkerboard artifacts due to the uneven overlapping. To address these limitations in one go, we propose a Multi-Domain Multi-Scale (MDMS) diffusion model for low-light image enhancement. In particular, we introduce a spatialfrequency fusion module to seamlessly integrates spatial and frequency information. By leveraging the Multi-Domain Learning (MDL) paradigm, our proposed model is endowed with the capability to adaptively facilitate noise distribution learning, thereby enhancing the quality of the generated images. Meanwhile, we propose a Multi-Scale Sampling (MSS) strategy that follows a divide-ensemble manner by merging the restored patches under different resolutions. Such a multi-scale learning paradigm explicitly derives patch information from different granularities, thus leading to smoother boundaries. Furthermore, we empirically adopt the Bright Channel Prior (BCP) which indicates natural statistical regularity as an additional restoration guidance. Experimental results on LOL and LOLv2 datasets demonstrate that our method achieves state-of-the-art performance for the low-light image enhancement task.

Please use this identifier to cite or link to this item:

http://hdl.handle.net/10453/185480