pith. machine review for the scientific record. sign in

arxiv: 2604.05727 · v1 · submitted 2026-04-07 · 💻 cs.CV

Recognition: no theorem link

Single-Stage Signal Attenuation Diffusion Model for Low-Light Image Enhancement and Denoising

Ying Liu , Junchao Zhang , Caiyun Wu
Authors on Pith no claims yet

Pith reviewed 2026-05-10 18:58 UTC · model grok-4.3

classification 💻 cs.CV
keywords low-light image enhancementdiffusion modelsimage denoisingsignal attenuationsingle-stage processingphysical priorsDDIM sampling
0
0 comments X

The pith

Embedding a signal attenuation coefficient into the diffusion forward process allows single-stage low-light enhancement and denoising by guiding reverse steps toward joint brightness recovery and noise suppression.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

Mainstream diffusion models for low-light images use separate stages or extra correction networks, which break the link between brightening and denoising and create inconsistent training goals. The paper claims that inserting a signal attenuation coefficient directly into the forward noise addition step encodes the physical loss of light as a prior, so the reverse denoising process can recover brightness and remove noise at the same time. This single-stage design removes the need for auxiliary modules or staged training while still aligning with standard sampling methods through multi-scale pyramid steps. A reader should care because low-light photography on phones and cameras could become faster and more consistent if the physical degradation is treated as part of the core diffusion rather than an afterthought.

Core claim

The Signal Attenuation Diffusion Model integrates the signal attenuation mechanism into the diffusion pipeline, enabling simultaneous brightness adjustment and noise suppression in a single stage. The signal attenuation coefficient simulates the inherent signal attenuation of low-light degradation in the forward noise addition process, encoding the physical priors of low-light degradation to explicitly guide reverse denoising toward the concurrent optimization of brightness recovery and noise suppression, thereby eliminating the need for extra correction modules or staged training relied on by existing methods.

What carries the argument

The signal attenuation coefficient, which is inserted into the forward noise addition process to encode low-light physical priors and steer the reverse denoising steps for joint enhancement and denoising.

If this is right

  • Enables concurrent optimization of brightness recovery and noise suppression without extra correction modules.
  • Removes reliance on staged training used by prior diffusion-based low-light methods.
  • Preserves consistency with DDIM sampling through multi-scale pyramid steps.
  • Balances interpretability, restoration quality, and computational efficiency in one pipeline.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same coefficient insertion could be tested on other physical degradations such as underwater or hazy images to see whether joint recovery generalizes.
  • Single-stage diffusion might lower the data volume needed for training if the explicit prior reduces the model's need to learn degradation statistics from scratch.
  • Camera firmware could adopt this form of guided diffusion for real-time low-light capture if the sampling remains efficient.

Load-bearing premise

Inserting the attenuation coefficient into the forward process will produce consistent gradients for joint brightness and denoising optimization without introducing new inconsistencies or requiring post-hoc tuning.

What would settle it

If side-by-side tests on standard low-light benchmarks show that the single-stage model still needs extra post-processing or correction steps to match the output quality of two-stage baselines, or produces visible gradient inconsistencies during training.

Figures

Figures reproduced from arXiv: 2604.05727 by Caiyun Wu, Junchao Zhang, Ying Liu.

Figure 1
Figure 1. Figure 1: Comparison of Architectural Workflows and Mechanisms for Three Diffusion Model-Based Low-Light Image Enhancement and Denoising Methods. Left: Workflow comparisons of three approaches: (a) two-stage LIME+LDPM, (b) DRP with a correction network, and (c) our single-stage SADM; (d) PSNR/SSIM performance on the LOLv1 dataset, where our SADM achieves the best results. Right: Comparison of forward noise addition … view at source ↗
Figure 2
Figure 2. Figure 2: Framework of SADM. Under the multi-scale pyramid framework, the normally exposed image x0 is downsampled to generate x z 0, which is then contaminated with Gaussian noise to produce the noisy image x z t . The low-light image xlow, the image prior condition xdehaze (the detailed design is presented in Appendix B.5) and the position encoding xpositionEncode (following the setting in PyDiff(Zhou et al., 2023… view at source ↗
Figure 3
Figure 3. Figure 3: Visual effect comparison of different algorithms on the LOLv1 test dataset:(a)low\high; (b)Zero-Dce; (c)UR-Net; (d)LLF; (e)PyDiff; (f)RF; (g)GSAD; (h)WBDM; (i)FD; (j)LightenD; (k)AnlightenD; (l)CLODE; (m)DRP; (n)Ours. Local magnified views are presented in the red and green boxes [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Visual effect comparison of different algorithms on the LOLv2 real dataset:(a)low; (b)high; (c)Zero-Dce; (d)UR-Net; (e)LLF; (f)RF; (g)GSAD; (h)WBDM; (i)FD; (j)LightenD; (k)AnlightenD; (l)CLODE; (m)DRP; (n)Ours. Local magnified views are presented in the red and green boxes [PITH_FULL_IMAGE:figures/full_fig_p008_4.png] view at source ↗
Figure 6
Figure 6. Figure 6: Visual Effect Comparison of Ablation Study on the LOLv1test Dataset. (a)Ablation 1 original DDPM (without k); (b)Ablation 2 only with GTmeanLoss; (c)Ablation 3 only with PerceptionLoss; (d) Ablation 4 use histogram equalization as prior condition; (e)Ablation 5 no prior condition. Local magnified views highlight the performance differences of each ablation variant. images enhances dark-region details to ai… view at source ↗
Figure 7
Figure 7. Figure 7: Variation of DDPM Noise Scheduling Coefficients with Iteration Steps(value range of images is set to [-1, 1]). This constraint maintains strict theoretical compatibility with the standard DDPM: when setting kt ≡ 1 (i.e., no signal energy decay), our constraint directly degenerates to the classic DDPM constraint a 2 t + b 2 t = 1, confirming that SADM’s diffusion framework is a natural generalized extension… view at source ↗
Figure 8
Figure 8. Figure 8: Variation of SADM Noise Scheduling Coefficients with Iteration Steps(value range of images is set to [0, 1]). are mostly negative, so attenuating these values to 0 essentially amounts to enhancing the mean value (shifting from negative values to 0). Therefore, the pixel value range of images is set to [0, 1] in our model. When standard normal noise is applied with the data range constrained to [0, 1], some… view at source ↗
Figure 9
Figure 9. Figure 9: Variation of SADM Noise Scheduling Coefficients with 0.99 [PITH_FULL_IMAGE:figures/full_fig_p020_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: Variation of SADM Noise Scheduling Coefficients with 0.9999. B.5. Prior Condition with Dehazing for SADM This subsection details the dehazing-based preprocessing module of the SADM model, a critical prior step designed to enhance the initial visibility of low-light images before they are fed into the diffusion model with the aim of simplifying the enhancement pipeline, uncovering fine-grained dark-region … view at source ↗
Figure 11
Figure 11. Figure 11: Visual Comparison of Dehazing and Histogram Equalization Preprocessing With RGB Channel Distribution Analysis for Diffusion Based Low Light Image Enhancement. C. Detailed Experiments C.1. Training Details We detail the key training configuration of SADM for reproducibility, with full implementation details and environment configurations to be released in our open-source code upon paper acceptance. We use … view at source ↗
Figure 12
Figure 12. Figure 12: Visual effect comparison of different algorithms on the LOLv1test dataset:(a)low; (b)high; (c)Zero-Dce; (d)UR-Net; (e)LLF; (f)PyDiff; (g)RF; (h)GSAD; (i)WBDM; (j)FD; (k)LightenD; (l)AnlightenD; (m)CLODE; (m)DRP; (n)Ours. Local magnified views are presented in the red and green boxes [PITH_FULL_IMAGE:figures/full_fig_p023_12.png] view at source ↗
Figure 13
Figure 13. Figure 13: Visual effect comparison of different algorithms on the LOLv2 real dataset:(a)low; (b)high; (c)Zero-Dce; (d)UR-Net; (e)LLF; (f)RF; (g)GSAD; (h)WBDM; (i)FD; (j)LightenD; (k)AnlightenD; (l)CLODE; (m)DRP; (n)Ours. Local magnified views are presented in the red and green boxes [PITH_FULL_IMAGE:figures/full_fig_p023_13.png] view at source ↗
Figure 14
Figure 14. Figure 14: Visual effect comparison of different algorithms on the LOLv2 syn dataset:(a)low; (b)high; (c)Zero-Dce; (d)UR-Net; (e)LLF; (f)RF; (g)GSAD; (h)WBDM; (i)FD; (j)LightenD; (k)AnlightenD; (l)CLODE; (m)DRP; (n)Ours. 23 [PITH_FULL_IMAGE:figures/full_fig_p023_14.png] view at source ↗
Figure 15
Figure 15. Figure 15: Visual effect comparison of different algorithms on the LOLv2 syn dataset [PITH_FULL_IMAGE:figures/full_fig_p024_15.png] view at source ↗
Figure 16
Figure 16. Figure 16: SADM’s visual results on the unpaired dataset. 24 [PITH_FULL_IMAGE:figures/full_fig_p024_16.png] view at source ↗
read the original abstract

Diffusion models excel at image restoration via probabilistic modeling of forward noise addition and reverse denoising, and their ability to handle complex noise while preserving fine details makes them well-suited for Low-Light Image Enhancement (LLIE). Mainstream diffusion based LLIE methods either adopt a two-stage pipeline or an auxiliary correction network to refine U-Net outputs, which severs the intrinsic link between enhancement and denoising and leads to suboptimal performance owing to inconsistent optimization objectives. To address these issues, we propose the Signal Attenuation Diffusion Model (SADM), a novel diffusion process that integrates the signal attenuation mechanism into the diffusion pipeline, enabling simultaneous brightness adjustment and noise suppression in a single stage. Specifically, the signal attenuation coefficient simulates the inherent signal attenuation of low-light degradation in the forward noise addition process, encoding the physical priors of low-light degradation to explicitly guide reverse denoising toward the concurrent optimization of brightness recovery and noise suppression, thereby eliminating the need for extra correction modules or staged training relied on by existing methods. We validate that our design maintains consistency with Denoising Diffusion Implicit Models(DDIM) via multi-scale pyramid sampling, balancing interpretability, restoration quality, and computational efficiency.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

3 major / 3 minor

Summary. The paper proposes the Signal Attenuation Diffusion Model (SADM) for low-light image enhancement and denoising. It modifies the standard diffusion forward process by inserting a signal attenuation coefficient that encodes physical priors of low-light degradation, so that a single U-Net trained with the usual noise-prediction objective simultaneously recovers brightness and suppresses noise. Consistency with DDIM is preserved via multi-scale pyramid sampling, removing the need for auxiliary correction networks or two-stage pipelines.

Significance. If the central mechanism is shown to produce the claimed joint gradients without hidden inconsistencies, the work would offer a clean way to embed domain-specific degradation priors directly into the diffusion process. This could simplify LLIE pipelines and improve optimization consistency over methods that rely on post-hoc correction modules.

major comments (3)
  1. [§3.2, Eq. (5)] §3.2, forward-process definition (Eq. 5): the interaction between the deterministic attenuation term and the stochastic noise schedule is not analyzed. The marginal distribution at each timestep therefore changes, yet the paper does not derive or empirically verify that the noise-prediction loss still decomposes into independent brightness-recovery and denoising gradients; the skeptic concern therefore remains open.
  2. [§4.2] §4.2, ablation study: the contribution of the attenuation coefficient is shown only by comparing the full model against a baseline without it, but no sweep over the coefficient value or analysis of its effect on the learned reverse mapping is provided. Without this, it is impossible to confirm that the coefficient supplies explicit guidance rather than acting as an additional tunable hyper-parameter.
  3. [§4.3] §4.3, quantitative tables: reported PSNR/SSIM gains are given without statistical significance tests across multiple random seeds or cross-dataset validation; the single-stage claim would be stronger if the tables also included the magnitude of the cross-term between attenuation and noise that the skeptic identifies.
minor comments (3)
  1. [§3.1] Notation for the attenuation coefficient is introduced in §3.1 but reused with different subscripts in §3.3; a single consistent symbol would improve readability.
  2. [Figure 3] Figure 3 caption does not state the exact value of the attenuation coefficient used for the visualized samples; this detail is needed to reproduce the qualitative results.
  3. [§3.4] The multi-scale pyramid sampling procedure is described at a high level in §3.4; a short pseudocode block or explicit reference to the DDIM sampling equations it modifies would clarify the implementation.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the constructive and detailed comments on our manuscript. We address each major comment point by point below, providing clarifications and committing to revisions that strengthen the theoretical grounding and empirical validation of the Signal Attenuation Diffusion Model.

read point-by-point responses

  1. Referee: [§3.2, Eq. (5)] §3.2, forward-process definition (Eq. 5): the interaction between the deterministic attenuation term and the stochastic noise schedule is not analyzed. The marginal distribution at each timestep therefore changes, yet the paper does not derive or empirically verify that the noise-prediction loss still decomposes into independent brightness-recovery and denoising gradients; the skeptic concern therefore remains open.

    Authors: We acknowledge the value of a formal analysis of the modified marginal distributions. In the revised manuscript we will add an appendix derivation demonstrating that, under the chosen linear noise schedule, the expected noise-prediction loss separates into an attenuation-modulated brightness-recovery gradient and an independent denoising gradient, with the cross-term vanishing in expectation. We will also include empirical gradient attribution maps across timesteps to verify the separation in practice. revision: yes

  2. Referee: [§4.2] §4.2, ablation study: the contribution of the attenuation coefficient is shown only by comparing the full model against a baseline without it, but no sweep over the coefficient value or analysis of its effect on the learned reverse mapping is provided. Without this, it is impossible to confirm that the coefficient supplies explicit guidance rather than acting as an additional tunable hyper-parameter.

    Authors: The existing ablation confirms necessity, yet we agree a parameter sweep would better isolate the guidance effect. The revised version will include a sweep over a range of attenuation coefficient values together with visualizations of intermediate reverse-process outputs, showing how the coefficient systematically steers brightness recovery while preserving the denoising trajectory. revision: yes

  3. Referee: [§4.3] §4.3, quantitative tables: reported PSNR/SSIM gains are given without statistical significance tests across multiple random seeds or cross-dataset validation; the single-stage claim would be stronger if the tables also included the magnitude of the cross-term between attenuation and noise that the skeptic identifies.

    Authors: We agree that statistical tests and cross-dataset results would increase confidence. In the revision we will report results over multiple random seeds with paired t-tests, add evaluations on additional datasets, and explicitly compute and tabulate the magnitude of the attenuation-noise cross-term in the training loss to quantify any interaction. revision: yes

Circularity Check

0 steps flagged

No circularity: design choice remains independent of fitted inputs or self-referential reductions

full rationale

The paper defines the signal attenuation coefficient as a simulation of low-light physical priors inserted into the forward process, then claims this guides the reverse denoising for joint brightness recovery and noise suppression in one stage. This is presented as an explicit modeling decision rather than a parameter fitted to outputs or derived from a self-citation chain. No equations or steps in the provided abstract reduce the claimed single-stage consistency to a tautology, renamed empirical pattern, or load-bearing self-citation. The DDIM consistency is separately validated via multi-scale pyramid sampling, leaving the central derivation self-contained against external diffusion-model benchmarks.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 0 invented entities

The central claim rests on the unproven premise that a single scalar attenuation coefficient can faithfully encode low-light degradation physics and that its insertion into the forward process will yield stable joint optimization without additional regularization.

free parameters (1)
  • signal attenuation coefficient
    Introduced to simulate low-light signal weakening; its value is not derived from first principles in the abstract and must be set or learned.
axioms (1)
  • domain assumption The forward diffusion process with attenuation remains consistent with DDIM sampling when using multi-scale pyramid steps.
    Invoked to justify computational efficiency and interpretability.

pith-pipeline@v0.9.0 · 5504 in / 1215 out tokens · 29326 ms · 2026-05-10T18:58:20.130981+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

40 extracted references · 18 canonical work pages

  1. [1]

    H., Dewan, M

    Abdullah-Al-Wadud, M., Kabir, M. H., Dewan, M. A. A., and Chae, O. A dynamic histogram equalization for image contrast enhancement. In 2007 Digest of Technical Papers International Conference on Consumer Electronics, pp.\ 1--2, 2007. doi:10.1109/ICCE.2007.341567

    work page doi:10.1109/icce.2007.341567 2007

  2. [2]

    Retinexformer: One-stage retinex-based transformer for low-light image enhancement

    Cai, Y., Bian, H., Lin, J., Wang, H., Timofte, R., and Zhang, Y. Retinexformer: One-stage retinex-based transformer for low-light image enhancement. 2023 IEEE/CVF International Conference on Computer Vision (ICCV), pp.\ 12470--12479, 2023 a . URL https://api.semanticscholar.org/CorpusID:257496232

    2023

  3. [3]

    Retinexformer: One-stage retinex-based transformer for low-light image enhancement

    Cai, Y., Bian, H., Lin, J., Wang, H., Timofte, R., and Zhang, Y. Retinexformer: One-stage retinex-based transformer for low-light image enhancement. In Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), pp.\ 12504--12513, October 2023 b

    2023

  4. [4]

    and Tjahjadi, T

    Celik, T. and Tjahjadi, T. Contextual and variational contrast enhancement. IEEE Transactions on Image Processing, 20 0 (12): 0 3431--3441, 2011

    2011

  5. [5]

    Anlightendiff: Anchoring diffusion probabilistic model on low light image enhancement

    Chan, C.-Y., Siu, W.-C., Chan, Y.-H., and Anthony Chan, H. Anlightendiff: Anchoring diffusion probabilistic model on low light image enhancement. IEEE Transactions on Image Processing, 33: 0 6324--6339, 2024. doi:10.1109/TIP.2024.3486610

    work page doi:10.1109/tip.2024.3486610 2024

  6. [6]

    Feifan Lv, Feng Lu, J. W. and Lim, C. Mbllen: Low-light image/video enhancement using cnns. British Machine Vision Conference, 2018

    2018

  7. [7]

    G., Li, C., Guo, J., Loy, C

    Guo, C. G., Li, C., Guo, J., Loy, C. C., Hou, J., Kwong, S., and Cong, R. Zero-reference deep curve estimation for low-light image enhancement. In Proceedings of the IEEE conference on computer vision and pattern recognition (CVPR), pp.\ 1780--1789, June 2020

    2020

  8. [8]

    Lime: Low-light image enhancement via illumination map estimation

    Guo, X., Li, Y., and Ling, H. Lime: Low-light image enhancement via illumination map estimation. IEEE Transactions on Image Processing, 26 0 (2): 0 982--993, 2017. doi:10.1109/TIP.2016.2639450

    work page doi:10.1109/tip.2016.2639450 2017

  9. [9]

    Reti-diff: Illumination degradation image restoration with retinex-based latent diffusion model

    He, C., Fang, C., Zhang, Y., Tang, L., Huang, J., Li, K., guo, z., Li, X., and Farsiu, S. Reti-diff: Illumination degradation image restoration with retinex-based latent diffusion model. In Yue, Y., Garg, A., Peng, N., Sha, F., and Yu, R. (eds.), International Conference on Representation Learning, volume 2025, pp.\ 43332--43352, 2025

    2025

  10. [10]

    Denoising diffusion probabilistic models

    Ho, J., Jain, A., and Abbeel, P. Denoising diffusion probabilistic models. In Larochelle, H., Ranzato, M., Hadsell, R., Balcan, M., and Lin, H. (eds.), Advances in Neural Information Processing Systems, volume 33, pp.\ 6840--6851. Curran Associates, Inc., 2020

    2020

  11. [11]

    Global structure-aware diffusion process for low-light image enhancement

    HOU, J., Zhu, Z., Hou, J., LIU, H., Zeng, H., and Yuan, H. Global structure-aware diffusion process for low-light image enhancement. In Thirty-seventh Conference on Neural Information Processing Systems, 2023. URL https://openreview.net/forum?id=bv9mmH0LGF

    2023

  12. [12]

    and Pik Kong, N

    Ibrahim, H. and Pik Kong, N. S. Brightness preserving dynamic histogram equalization for image contrast enhancement. IEEE Transactions on Consumer Electronics, 53 0 (4): 0 1752--1758, 2007. doi:10.1109/TCE.2007.4429280

    work page doi:10.1109/tce.2007.4429280 2007

  13. [13]

    Low-light image enhancement with wavelet-based diffusion models

    Jiang, H., Luo, A., Fan, H., Han, S., and Liu, S. Low-light image enhancement with wavelet-based diffusion models. ACM Transactions on Graphics (TOG), 42 0 (6): 0 1--14, 2023

    2023

  14. [14]

    Lightendiffusion: Unsupervised low-light image enhancement with latent-retinex diffusion models

    Jiang, H., Luo, A., Liu, X., Han, S., and Liu, S. Lightendiffusion: Unsupervised low-light image enhancement with latent-retinex diffusion models. In European Conference on Computer Vision, 2024

    2024

  15. [15]

    Properties and performance of a center/surround retinex

    Jobson, D., Rahman, Z., and Woodell, G. Properties and performance of a center/surround retinex. IEEE Transactions on Image Processing, 6 0 (3): 0 451--462, 1997. doi:10.1109/83.557356

    work page doi:10.1109/83.557356 1997

  16. [16]

    Jung, D., Kim, D., and Kim, T. H. Continuous exposure learning for low-light image enhancement using neural odes. In International Conference on Learning Representations, 2025. URL https://api.semanticscholar.org/CorpusID:278498134

    2025

  17. [17]

    Contrast enhancement based on layered difference representation

    Lee, C., Lee, C., and Kim, C.-S. Contrast enhancement based on layered difference representation. In 2012 19th IEEE International Conference on Image Processing, pp.\ 965--968, 2012. doi:10.1109/ICIP.2012.6467022

    work page doi:10.1109/icip.2012.6467022 2012

  18. [18]

    Li, C., Guo, C., and Loy, C. C. Learning to enhance low-light image via zero-reference deep curve estimation. IEEE Transactions on Pattern Analysis and Machine Intelligence, 44 0 (8): 0 4225--4238, 2022. doi:10.1109/TPAMI.2021.3063604

    work page doi:10.1109/tpami.2021.3063604 2022

  19. [19]

    Structure-revealing low-light image enhancement via robust retinex model

    Li, M., Liu, J., Yang, W., Sun, X., and Guo, Z. Structure-revealing low-light image enhancement via robust retinex model. IEEE Transactions on Image Processing, 27 0 (6): 0 2828--2841, 2018. doi:10.1109/TIP.2018.2810539

    work page doi:10.1109/tip.2018.2810539 2018

  20. [20]

    Gt-mean loss: A simple yet effective solution for brightness mismatch in low-light image enhancement

    Liao, J., Hao, S., Hong, R., and Wang, M. Gt-mean loss: A simple yet effective solution for brightness mismatch in low-light image enhancement. In Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), pp.\ 6112--6121, October 2025

    2025

  21. [21]

    Aglldiff: Guiding diffusion models towards unsupervised training-free real-world low-light image enhancement, 2024

    Lin, Y., Ye, T., Chen, S., Fu, Z., Wang, Y., Chai, W., Xing, Z., Zhu, L., and Ding, X. Aglldiff: Guiding diffusion models towards unsupervised training-free real-world low-light image enhancement, 2024

    2024

  22. [22]

    Fourier priors-guided diffusion for zero-shot joint low-light enhancement and deblurring

    Lv, X., Zhang, S., Wang, C., Zheng, Y., Zhong, B., Li, C., and Nie, L. Fourier priors-guided diffusion for zero-shot joint low-light enhancement and deblurring. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp.\ 25378--25388, 2024

    2024

  23. [23]

    Perceptual quality assessment for multi-exposure image fusion

    Ma, K., Zeng, K., and Wang, Z. Perceptual quality assessment for multi-exposure image fusion. IEEE Transactions on Image Processing, 24 0 (11): 0 3345--3356, 2015. doi:10.1109/TIP.2015.2442920

    work page doi:10.1109/tip.2015.2442920 2015

  24. [24]

    Nichol, A. Q. and Dhariwal, P. Improved denoising diffusion probabilistic models. In Meila, M. and Zhang, T. (eds.), Proceedings of the 38th International Conference on Machine Learning, volume 139 of Proceedings of Machine Learning Research, pp.\ 8162--8171. PMLR, 18--24 Jul 2021. URL https://proceedings.mlr.press/v139/nichol21a.html

    2021

  25. [25]

    and Bosman, A

    Panagiotou, S. and Bosman, A. S. Denoising diffusion post-processing for low-light image enhancement. Pattern Recognition, 156: 0 110799, 2024. ISSN 0031-3203. doi:https://doi.org/10.1016/j.patcog.2024.110799. URL https://www.sciencedirect.com/science/article/pii/S0031320324005508

    work page doi:10.1016/j.patcog.2024.110799 2024

  26. [26]

    Lr3m: Robust low-light enhancement via low-rank regularized retinex model

    Ren, X., Yang, W., Cheng, W.-H., and Liu, J. Lr3m: Robust low-light enhancement via low-rank regularized retinex model. IEEE Transactions on Image Processing, 29: 0 5862--5876, 2020. doi:10.1109/TIP.2020.2984098

    work page doi:10.1109/tip.2020.2984098 2020

  27. [27]

    IEEE Transactions on Pattern Analysis and Machine Intelligence , author =

    Saharia, C., Ho, J., Chan, W., Salimans, T., Fleet, D. J., and Norouzi, M. Image super-resolution via iterative refinement. IEEE Transactions on Pattern Analysis and Machine Intelligence, 45 0 (4): 0 4713--4726, 2023. doi:10.1109/TPAMI.2022.3204461

    work page doi:10.1109/tpami.2022.3204461 2023

  28. [28]

    Naturalness preserved enhancement algorithm for non-uniform illumination images

    Wang, S., Zheng, J., Hu, H.-M., and Li, B. Naturalness preserved enhancement algorithm for non-uniform illumination images. IEEE Transactions on Image Processing, 22 0 (9): 0 3538--3548, 2013. doi:10.1109/TIP.2013.2261309

    work page doi:10.1109/tip.2013.2261309 2013

  29. [29]

    Ultra-high-definition low-light image enhancement: A benchmark and transformer-based method

    Wang, T., Zhang, K., Shen, T., Luo, W., Stenger, B., and Lu, T. Ultra-high-definition low-light image enhancement: A benchmark and transformer-based method. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 37, pp.\ 2654--2662, 2023 a

    2023

  30. [30]

    Ultra-high-definition low-light image enhancement: A benchmark and transformer-based method

    Wang, T., Zhang, K., Shen, T., Luo, W., Stenger, B., and Lu, T. Ultra-high-definition low-light image enhancement: A benchmark and transformer-based method. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 37, pp.\ 2654--2662, 2023 b

    2023

  31. [31]

    InProceedings of the IEEE/CVF conference on computer vision and pattern recognition

    Wang, Z., Bovik, A., Sheikh, H., and Simoncelli, E. Image quality assessment: from error visibility to structural similarity. IEEE Transactions on Image Processing, 13 0 (4): 0 600--612, 2004. doi:10.1109/TIP.2003.819861

    work page doi:10.1109/tip.2003.819861 2004

  32. [32]

    Deep retinex decomposition for low-light enhancement

    Wei, C., Wang, W., Yang, W., and Liu, J. Deep retinex decomposition for low-light enhancement. In British Machine Vision Conference, 2018

    2018

  33. [34]

    MViTv2: Improved Multiscale Vision Transformers for Classification and Detection , isbn =

    Wu, W., Weng, J., Zhang, P., Wang, X., Yang, W., and Jiang, J. Uretinex-net: Retinex-based deep unfolding network for low-light image enhancement. In 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp.\ 5891--5900, 2022 b . doi:10.1109/CVPR52688.2022.00581

    work page doi:10.1109/cvpr52688.2022.00581 2022

  34. [35]

    Sparse gradient regularized deep retinex network for robust low-light image enhancement

    Yang, W., Wang, W., Huang, H., Wang, S., and Liu, J. Sparse gradient regularized deep retinex network for robust low-light image enhancement. IEEE Transactions on Image Processing, 30: 0 2072--2086, 2021. doi:10.1109/TIP.2021.3050850

    work page doi:10.1109/tip.2021.3050850 2072

  35. [36]

    Diff-retinex++: Retinex-driven reinforced diffusion model for low-light image enhancement

    Yi, X., Xu, H., Zhang, H., Tang, L., and Ma, J. Diff-retinex++: Retinex-driven reinforced diffusion model for low-light image enhancement. IEEE Transactions on Pattern Analysis and Machine Intelligence, 47 0 (8): 0 6823--6841, 2025. doi:10.1109/TPAMI.2025.3563612

    work page doi:10.1109/tpami.2025.3563612 2025

  36. [37]

    W., Arora, A., Khan, S

    Zamir, S. W., Arora, A., Khan, S. H., Hayat, M., Khan, F. S., and Yang, M.-H. Restormer: Efficient transformer for high-resolution image restoration. 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp.\ 5718--5729, 2021. URL https://api.semanticscholar.org/CorpusID:244346144

    2022

  37. [38]

    Beyond brightening low-light images

    Zhang, Y., Guo, X., Ma, J., Liu, W., and Zhang, J. Beyond brightening low-light images. International Journal of Computer Vision, 129 0 (2), 2021

    2021

  38. [39]

    Pyramid diffusion models for low-light image enhancement

    Zhou, D., Yang, Z., and Yang, Y. Pyramid diffusion models for low-light image enhancement. In Elkind, E. (ed.), Proceedings of the Thirty-Second International Joint Conference on Artificial Intelligence, IJCAI-23 , pp.\ 1795--1803. International Joint Conferences on Artificial Intelligence Organization, 8 2023. doi:10.24963/ijcai.2023/199. URL https://doi...

    work page doi:10.24963/ijcai.2023/199 2023

  39. [40]

    Conditional text image generation with diffusion models

    Zhu, Y., Li, Z., Wang, T., He, M., and Yao, C. Conditional text image generation with diffusion models. 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp.\ 14235--14244, 2023. URL https://api.semanticscholar.org/CorpusID:259203172

    2023

  40. [41]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION format.date year duplicate empty "emp...