Zero-Reference Lighting Estimation Diffusion Model for Low-Light Image Enhancement

Jinhong He, Minglong Xue, Aoxiang Ning, Chengyun Song
Proceedings of the 16th Asian Conference on Machine Learning, PMLR 260:1160-1175, 2025.

Abstract

Diffusion model-based low-light image enhancement methods rely heavily on paired training data, which limits its extensive application. Meanwhile, existing unsupervised methods lack effective bridging capabilities for unknown degradation. To address these limitations, we firstly propose a novel zero-reference lighting estimation diffusion model for low-light image enhancement called Zero-LED. It utilizes the stable convergence ability of diffusion models to bridge the gap between low-light domains and real normal-light domains and successfully alleviates the dependence on pairwise training data via zero-reference learning. Specifically, we first design the initial optimization network to preprocess the input image and implement bidirectional constraints between the diffusion model and the initial optimization network through multiple objective functions. Subsequently, the degradation factors of the real-world scene are optimized iteratively to achieve effective light enhancement. In addition, we explore a frequency-domain based and semantically guided appearance reconstruction module that encourages feature alignment of the recovered image at a fine-grained level and satisfies subjective expectations. Finally, extensive experiments demonstrate the superiority of our approach to other state-of-the-art methods and more significant generalization capabilities.

Cite this Paper

BibTeX
@InProceedings{pmlr-v260-he25b, title = {Zero-Reference Lighting Estimation Diffusion Model for Low-Light Image Enhancement}, author = {He, Jinhong and Xue, Minglong and Ning, Aoxiang and Song, Chengyun}, booktitle = {Proceedings of the 16th Asian Conference on Machine Learning}, pages = {1160--1175}, year = {2025}, editor = {Nguyen, Vu and Lin, Hsuan-Tien}, volume = {260}, series = {Proceedings of Machine Learning Research}, month = {05--08 Dec}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v260/main/assets/he25b/he25b.pdf}, url = {https://proceedings.mlr.press/v260/he25b.html}, abstract = {Diffusion model-based low-light image enhancement methods rely heavily on paired training data, which limits its extensive application. Meanwhile, existing unsupervised methods lack effective bridging capabilities for unknown degradation. To address these limitations, we firstly propose a novel zero-reference lighting estimation diffusion model for low-light image enhancement called Zero-LED. It utilizes the stable convergence ability of diffusion models to bridge the gap between low-light domains and real normal-light domains and successfully alleviates the dependence on pairwise training data via zero-reference learning. Specifically, we first design the initial optimization network to preprocess the input image and implement bidirectional constraints between the diffusion model and the initial optimization network through multiple objective functions. Subsequently, the degradation factors of the real-world scene are optimized iteratively to achieve effective light enhancement. In addition, we explore a frequency-domain based and semantically guided appearance reconstruction module that encourages feature alignment of the recovered image at a fine-grained level and satisfies subjective expectations. Finally, extensive experiments demonstrate the superiority of our approach to other state-of-the-art methods and more significant generalization capabilities.} }
Endnote
%0 Conference Paper %T Zero-Reference Lighting Estimation Diffusion Model for Low-Light Image Enhancement %A Jinhong He %A Minglong Xue %A Aoxiang Ning %A Chengyun Song %B Proceedings of the 16th Asian Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2025 %E Vu Nguyen %E Hsuan-Tien Lin %F pmlr-v260-he25b %I PMLR %P 1160--1175 %U https://proceedings.mlr.press/v260/he25b.html %V 260 %X Diffusion model-based low-light image enhancement methods rely heavily on paired training data, which limits its extensive application. Meanwhile, existing unsupervised methods lack effective bridging capabilities for unknown degradation. To address these limitations, we firstly propose a novel zero-reference lighting estimation diffusion model for low-light image enhancement called Zero-LED. It utilizes the stable convergence ability of diffusion models to bridge the gap between low-light domains and real normal-light domains and successfully alleviates the dependence on pairwise training data via zero-reference learning. Specifically, we first design the initial optimization network to preprocess the input image and implement bidirectional constraints between the diffusion model and the initial optimization network through multiple objective functions. Subsequently, the degradation factors of the real-world scene are optimized iteratively to achieve effective light enhancement. In addition, we explore a frequency-domain based and semantically guided appearance reconstruction module that encourages feature alignment of the recovered image at a fine-grained level and satisfies subjective expectations. Finally, extensive experiments demonstrate the superiority of our approach to other state-of-the-art methods and more significant generalization capabilities.
APA
He, J., Xue, M., Ning, A. & Song, C.. (2025). Zero-Reference Lighting Estimation Diffusion Model for Low-Light Image Enhancement. Proceedings of the 16th Asian Conference on Machine Learning, in Proceedings of Machine Learning Research 260:1160-1175 Available from https://proceedings.mlr.press/v260/he25b.html.

Related Material