Pith. sign in

REVIEW 4 major objections 5 minor 2 cited by

Zero-Shot Low Light Image Enhancement with Diffusion Prior

T0 review · 4 major / 5 minor · reviewed 2026-08-11 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read A frozen text-to-image diffusion model, with no optimization or per-image tuning, can enhance low-light images by guiding denoising with its own self-attention features.

desk verdict A useful zero-shot diffusion pipeline with clean ablations, but the 'free lunch' claim is undercut by the borrowed QuadPrior decoder. read the letter →

arxiv 2412.13401 v4 pith:CRJ2BGQK submitted 2024-12-18 cs.CV

Joshua Cho , Sara Aghajanzadeh , Zhen Zhu , D. A. Forsyth This is my paper · ORCID
classification cs.CV
keywords low-lightimageenhancementzero-shotdiffusionpriorself-attentionautowhitebalanceDDIMinversioncolorconstancyAdaIN
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper argues that a frozen text-to-image diffusion model, used as-is, already has enough internal structure to enhance low-light images without any per-image optimization, training, tuning, or text prompt. The proposed procedure inverts the input into a noisy latent, recenters that latent's channel statistics onto a standard normal distribution, then denoises while swapping in the self-attention features extracted from the input during inversion. In experiments on paired and unpaired benchmarks, the authors report that this zero-shot procedure outperforms existing unsupervised and zero-shot methods on most quality metrics and lands close to supervised methods; the same unmodified procedure also produces competitive auto white balance results. The larger suggestion is that a diffusion model's own features can replace task-specific degradation assumptions across multiple color-recovery tasks.

What carries the argument

The central machinery is a four-step, training-free pipeline. Step 1 lifts the average intensity of a very dark input to a threshold of 30. Step 2 applies DDIM inversion, which maps the input to a noised latent $z^c_T$ and records the self-attention features $\{q^l_t, k^l_t, v^l_t\}$ from each up-block layer of the diffusion U-Net across timesteps. Step 3 applies Adaptive Instance Normalization (AdaIN), a channel-wise statistics-matching operation, to re-center the inverted latent on a standard normal distribution: $z^*_T = \sigma(z^s_T) \frac{z^c_T - \mu(z^c_T)}{\sigma(z^c_T)} + \mu(z^s_T)$, with $z^s_T \sim N(0,I)$. Step 4 denoises $z^*_T$ by DDIM sampling while replacing the default self-attention with the recorded features, which prevents the drift that direct DDIM denoising would otherwise introduce. Self-attention features are the query, key, and value tensors that let each spatial location of the latent attend to every other location; they encode layout and semantics and are largely invariant to illumination. The load-bearing hypothesis is that the VAE latent channels are loosely aligned with color properties, so matching their statistics to a standard normal shifts a dark, color-shifted latent toward natural, well-lit colors.

What would settle it

Take a single low-light scene, render it under two different color temperatures so that the same surfaces have different casts, and run the pipeline on both; if the two outputs do not converge toward the same surface colors, the color constancy claim is false. For a more direct test of the color-alignment hypothesis, construct an image with a known linear color cast, enhance it, and check whether the channel-wise means of the AdaIN-adjusted latent shift along the inverse-cast direction; if they do not, the recentering is not doing the claimed color work.

Watch

Extended reading notes

Core claim

The paper's central claim is that a pre-trained text-to-image diffusion model, with no optimization, training, fine-tuning, text conditioning, or hyperparameter adjustment, can consistently reconstruct low-light images with high fidelity, and that the identical unmodified procedure performs auto white balance at a level comparable to supervised methods. The mechanism is to let the input image's own self-attention features carry the structure: DDIM inversion moves the image to a noised latent and records the attention maps, AdaIN shifts the latent's channel-wise mean and standard deviation to match a standard normal draw, and DDIM denoising replaces the model's default self-attention with the recorded features so the output stays tied to the input. The authors present quantitative results on LOL, LSRW, five unpaired datasets, and the CUBE+ white-balance benchmark, and they interpret the procedure as a 'free lunch' alternative to zero-shot methods that optimize per-image parameters.

Load-bearing premise

The load-bearing premise is that the VAE's latent channels align with color properties closely enough that recentering the inverted latent's statistics onto a standard normal distribution moves the decoded image from dark and color-shifted toward natural colors, and that self-attention features stay stable enough under changing illumination to guide the denoising.

Editorial extensions

If this is right

  • Because no per-image optimization is needed, the added inference cost over a single diffusion pass is essentially one inversion pass plus one sampling pass, compared with per-image parameter updates in GDP, TAO, and FourierDiff.
  • The same unmodified pipeline can be applied to any color-imbalanced image, as demonstrated by the paper's auto white balance experiments on CUBE+.
  • The quality ceiling is set by the pretrained diffusion model's natural-image prior, so the method should generalize to scenes and lighting conditions not seen by any LLIE training set.
  • Because self-attention guidance comes from the input itself, the method avoids the hallucinations the paper attributes to per-image optimized diffusion baselines, at least on the tested datasets.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • Not claimed by the paper: the same AdaIN recentering could be reused as a drop-in color-correction module in other latent-diffusion tasks, such as underwater image correction or stylization, wherever the VAE channel–color alignment holds.
  • Not claimed by the paper: if the channel–color alignment is an artifact of this particular VAE/decoder pair, the method's color correction may not transfer to other latent diffusion models; a direct test would be running the identical pipeline with a different pretrained VAE and checking whether the color shift is preserved.
  • Not claimed by the paper: applying the pipeline frame-by-frame to video would likely produce flicker because self-attention features are extracted per frame with no temporal consistency mechanism.
  • Not claimed by the paper: the method's robustness to inputs with very high brightness is demonstrated qualitatively, but the paper does not characterize the operating range of the preprocessing threshold; a stress test on overexposed inputs would map its boundaries.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 5 minor

Summary. The paper proposes a zero-shot method for low-light image enhancement and auto white balance. The pipeline is: preprocess by scaling up the average intensity if it is below a threshold of 30, invert the preprocessed input with DDIM inversion using T=25 steps, apply AdaIN to the inverted latent to match N(0,I), denoise while replacing the self-attention features in the up blocks with features extracted during inversion, and finally decode with the VAE decoder taken from QuadPrior. The authors claim that this requires no optimization, training, fine-tuning, text conditioning, or hyperparameter adjustments, and that it outperforms existing unsupervised and zero-shot methods on standard benchmarks while matching supervised methods. Quantitative results are reported on LOL, LSRW, several unpaired datasets, and the CUBE+ AWB dataset.

Significance. If the central claim were fully supported, this would be a valuable result: a training-free, optimization-free inference procedure that turns a frozen text-to-image diffusion model into a low-light enhancement and white-balance tool would be conceptually clean and practically attractive. The paper has genuine strengths: the ablation study in Table 3 isolates several components (self-attention replacement, AdaIN, decoder choice, preprocessing level), the dataset counts are given precisely, and the AWB evaluation uses an established protocol. However, the main scientific claim is weakened by the use of a decoder borrowed from QuadPrior, an unsupervised low-light enhancement method, and by hand-set parameters that are presented as absent. The paper is therefore best viewed as a promising empirical pipeline whose attribution of the gains to the frozen diffusion prior alone needs substantially more evidence.

major comments (4)
  1. [Section 3 (Ours w/ SD Decoder) and Table 1 footnote] The final pipeline decodes with the VAE decoder from QuadPrior [76], which is itself an unsupervised low-light enhancement method trained on COCO. Table 3 shows that replacing this decoder with the default Stable Diffusion decoder lowers PSNR on LOL from 21.739 to 19.927 and SSIM from 0.815 to 0.600, a large drop. This means that a trained, task-specific component contributes substantially to the reported results, so the advertised 'free lunch' improvement cannot be attributed solely to the frozen diffusion prior and self-attention injection. The same confound applies to the AWB results in Table 2. The authors should either report a controlled comparison using the QuadPrior decoder with a simpler enhancement baseline, or explicitly state that the method includes a trained LLIE decoder and adjust the 'no training' claim accordingly.
  2. [Abstract and Section 3 (Preprocessing, Inversion)] The claim of 'no hyperparameter adjustments' is contradicted by the hand-set average-intensity threshold of 30.0 and the hand-set DDIM inversion step count T=25. Table 3 also shows sensitivity to the preprocessing choice: 'Ours w/ SA avg=60' is worse than the final setting, and 'Ours w/ SA avg=input' is also different. The paper should present a sensitivity analysis for these two parameters and replace the 'no hyperparameters' claim with a more precise statement such as 'uses fixed default hyperparameters.'
  3. [Abstract and Section 4.1] The statements that the approach 'consistently reconstructs low-light images with superior fidelity' and 'outperforms SOTA methods' are too broad. In Table 1, on LOL the supervised method SNR-Aware [80] achieves PSNR 21.8877 compared to the reported 21.7393 for the proposed method, and on LSRW Diff-Plugin [46] achieves PSNR 17.9620 versus 17.6634. The manuscript's own quantitative results therefore support a more qualified claim, namely that the method is competitive with or better than existing zero-shot and unsupervised methods on several metrics and datasets, not that it consistently surpasses all state-of-the-art methods.
  4. [Section 5] The explanatory mechanism is presented as two hypotheses: that latent channels align with color properties, and that self-attention features are largely invariant to illumination and white balance. These hypotheses are not independently validated, and Table 3 indicates that self-attention is not fully invariant: 'Ours w/ SA avg=60' degrades relative to the final setting. The paper should either provide direct evidence for the channel-color alignment (for example, by analyzing which channels change after AdaIN and how that affects output color statistics) or present the explanation more cautiously as a plausible interpretation rather than a demonstrated cause.
minor comments (5)
  1. [Table 1 caption] The caption reads 'Qualitative comparison' but the table reports quantitative metrics; it should be 'Quantitative comparison.'
  2. [Table 2 text and References] The text says the method outperforms 'DDIM [69]' but the corresponding row is DDNM [77]; also, reference [73] is incomplete ('Timothy Alexis Vass, 2024') and should be completed.
  3. [Table 3] The variant name 'Ours w/ SA (DDIM Sampling)' is confusing because all variants use DDIM sampling; it should clarify that the variant extracts self-attention during sampling rather than during inversion.
  4. [Figure 9] The label 'AdaIN + DDIM + SADDIM' appears to contain a typo or an undefined abbreviation; please use a consistent, defined naming scheme.
  5. [Figure 8 caption] The caption is very long and partly repeats the main text; it would be easier to read if the key claim about channel-color alignment were stated once and the figure focused on the visual evidence.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: fixed frozen-pipeline composition, no fitted parameters.

full rationale

The derivation chain is self-contained and not circular. The method is a fixed composition of four operations: a hand-set intensity floor of 30.0 in preprocessing, frozen DDIM inversion of a VAE-encoded latent, a deterministic AdaIN recentering of the inverted latent to N(0,I), and denoising with self-attention features captured during inversion. No parameter is fitted to the evaluation benchmarks, and the reported PSNR, SSIM, LPIPS, and DeltaE values are measured outputs rather than quantities defined by the inputs. The strongest circularity candidate is the reuse of QuadPrior's VAE decoder, a zero-reference LLIE method trained on COCO; however, the paper discloses this in the Table 1 footnote (As we leverage decoder component of VAE from QuadPrior [76]) and explicitly ablates it in Section 5: replacing the QuadPrior decoder with the default Stable Diffusion decoder changes LOL PSNR from 21.739 to 19.927, and removing self-attention guidance drops PSNR to 13.173. Thus the reported result does not reduce to the borrowed decoder alone; this is a borrowing/attribution concern that weakens the free-lunch framing but is not a reasoning loop. The latent-channel color-alignment idea is explicitly framed as a hypothesis in Section 5 (we hypothesize that each channel predominantly aligns with a specific color property), not as a derived theorem. No equation equates target metrics to inputs by construction, and no load-bearing argument rests on the authors' prior self-citations.

Assumptions & free parameters 3 free parameters · 5 assumptions · 0 invented entities

The central method has no fitted numerical parameters in the classical sense, but two hand-set constants and four unproven model assumptions are load-bearing. The QuadPrior decoder is an external task-specific component, not an invented entity. No new particles, forces, or model entities are introduced.

free parameters (3)
  • average intensity threshold = 30.0 on a 0-255 scale
    Input images whose average intensity is below 30 are scaled up to 30 before inversion; the paper reports no sensitivity analysis across datasets, and the Table 3 ablation at threshold 60 changes LPIPS from 0.177 to 0.194.
  • DDIM inversion steps T = 25
    The inversion and sampling use T=25; no ablation on T is reported, and inversion error at small T is known to be nontrivial.
  • self-attention layers used for replacement = up blocks only
    Only up-block self-attention is replaced; down-block and cross-attention treatment is unspecified, a design choice that affects the output.
assumptions (5)
  • domain assumption Stable Diffusion's pretrained prior is a suitable natural-image prior for low-light images despite being trained on well-lit images.
    The whole pipeline assumes the SD latent space supports meaningful denoising from darkened latents (Sec. 3).
  • domain assumption DDIM inversion with T=25 produces self-attention features that faithfully represent the input's semantics and are usable as guidance during sampling.
    Inversion errors are acknowledged in Sec. 5, and self-attention replacement is the mitigation; if the inversion trajectory is poor, guidance is poor.
  • ad hoc to paper Self-attention features are largely invariant to illumination and white balance, so input-derived attention can guide a well-lit reconstruction.
    Stated in Sec. 5 as the basis for attention replacement; not proven by controlled experiments.
  • ad hoc to paper Latent channels predominantly align with specific color properties, so AdaIN recentering to N(0,I) corrects color shifts.
    Explicitly labeled a hypothesis in Sec. 5 and Figure 8; load-bearing for the color-constancy claim.
  • domain assumption QuadPrior's VAE decoder reconstructs latents with less distortion than the SD decoder for this task.
    The paper switches decoders for robust self-reconstruction (Sec. 5) without reporting the comparison that justifies this claim; it also imports a component from an LLIE baseline.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Zero-Shot Low Light Image Enhancement with Diffusion Prior." pith.science (2026). https://pith.science/paper/CRJ2BGQK

@misc{pith2026241213401,
  author       = {Pith},
  title        = {Pith review of: Zero-Shot Low Light Image Enhancement with Diffusion Prior},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/CRJ2BGQK}},
  note         = {Machine review of arXiv:2412.13401}
}
read the original abstract

In this paper, we present a simple yet highly effective "free lunch" solution for low-light image enhancement (LLIE), which aims to restore low-light images as if acquired in well-illuminated environments. Our method necessitates no optimization, training, fine-tuning, text conditioning, or hyperparameter adjustments, yet it consistently reconstructs low-light images with superior fidelity. Specifically, we leverage a pre-trained text-to-image diffusion prior, learned from training on a large collection of natural images, and the features present in the model itself to guide the inference, in contrast to existing methods that depend on customized constraints. Comprehensive quantitative evaluations demonstrate that our approach outperforms SOTA methods on established datasets, while qualitative analyses indicate enhanced color accuracy and the rectification of subtle chromatic deviations. Furthermore, additional experiments reveal that our method, without any modifications, achieves SOTA-comparable performance in the auto white balance (AWB) task.

Figures

Figures reproduced from arXiv: 2412.13401 by the authors.

Figure 1
Figure 1. In low-light image enhancement, an ideal method should achieve [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Hallucination. While diffusion prior is effective for im￾age restoration, improper application can lead to unintended hal￾lucinations, where the model generates nonexistent structures or alters scene semantics. For example, GDP [20], a robust and ver￾satile image restoration method, often hallucinates in the presence of substantial noise and darkness in input images. As shown in row 1, a blue-colored cabinet is inac… view at source ↗
Figure 4
Figure 4. Overall pipeline of our method. Our method offers a simple yet highly effective “free lunch” solution for both LLIE and AWB and consists of four main steps: (1) preprocessing; (2) inverting the input image; (3) adjusting the resulting noised latent with Adaptive Instance Normalization (AdaIN) to match standard normal distributions N (0, I); and (4) denoising the inverted representation with self￾attention features e… view at source ↗
Figures from the paper (6 more)
Figure 5
Figure 5. Figure 5: Qualitative evaluation of our method against existing unsupervised and zero-shot approaches on the paired LOL dataset. Please [PITH_FULL_IMAGE:figures/full_fig_p006_5.png]
Figure 6
Figure 6. Figure 6: Qualitative evaluation of our method against existing unsupervised, supervised, and zero-shot approaches on the Unpaired dataset. [PITH_FULL_IMAGE:figures/full_fig_p006_6.png]
Figure 7
Figure 7. Figure 7: Qualitative evaluation of our method against existing supervised and general image restoration methods on auto white balance [PITH_FULL_IMAGE:figures/full_fig_p007_7.png]
Figure 8
Figure 8. Figure 8: The top row shows the average RGB histograms of 100 randomly selected LOL images for each channel (0, 1, 2, 3) when [PITH_FULL_IMAGE:figures/full_fig_p007_8.png]
Figure 10
Figure 10. Figure 10: Failure cases. Our method maintains color channel values centered around 100. While this characteristic may occa￾sionally lead to deviations from the ground truth, it proves advan￾tageous in most cases, particularly for low-light images where sig￾nificant color inform…
Figure 9
Figure 9. Figure 9: Qualitative ablation study. The images in this figure are sourced from the LOL dataset. Please refer to Section 5 for a detailed discussion. Prior [76], which demonstrates superior self-reconstruction compared to the default Stable Diffusion decoder [63], of￾ten prone …

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 2 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Taming Stable Diffusion for Computed Tomography Blind Super-Resolution

    eess.IV 2025-06 conditional novelty 6.0 of 10

    A Stable Diffusion framework with LLaVA-Med text prompts and a side-controlling network achieves state-of-the-art CT blind super-resolution on two datasets under synthetic degradation.

  2. High-Quality Exposure Correction with Diffusion-Based Image Generation Priors

    cs.CV 2026-08 conditional novelty 5.0 of 10

    DPEC fine-tunes a pretrained diffusion model for single-step exposure correction and fuses its low-frequency output into a regression network, improving perceptual metrics on LCDP, MSEC, and SICE.

Reference graph

Works this paper leans on

92 extracted references · 79 canonical work pages · cited by 2 Pith papers

  1. [76]

    Zero-reference low-light enhancement via physical quadru- ple priors

    Wenjing Wang, Huan Yang, Jianlong Fu, and Jiaying Liu. Zero-reference low-light enhancement via physical quadru- ple priors. In CVPR, 2024. 3, 5, 8

    work page 2024

  2. [80]

    Snr-aware low-light image enhancement

    Xiaogang Xu, Ruixing Wang, Chi-Wing Fu, and Jiaya Jia. Snr-aware low-light image enhancement. In CVPR, 2022. 5

    work page 2022

  3. [46]

    Yuhao Liu, Zhanghan Ke, Fang Liu, Nanxuan Zhao, and Rynson W.H. Lau. Diff-plugin: Revitalizing details for diffusion-based low-level tasks. In CVPR, 2024. 3, 5

    work page 2024

  4. [1]

    When color constancy goes wrong: Correcting im- properly white-balanced images

    Mahmoud Afifi, Brian Price, Scott Cohen, and Michael S Brown. When color constancy goes wrong: Correcting im- properly white-balanced images. In CVPR, 2019. 3, 5, 7, 8

    2019

  5. [2]

    Brubaker, and Michael S

    Mahmoud Afifi, Marcus A. Brubaker, and Michael S. Brown. Auto white-balance correction for mixed-illuminant scenes. In WACV, 2022. 3, 5, 7, 8

    2022

  6. [3]

    Towards Robust Low Light Image Enhancement

    Sara Aghajanzadeh and David Forsyth. Towards robust low light image enhancement. arXiv:2205.08615, 2022. 3

    work page Pith review arXiv 2022

  7. [4]

    Long Scale Error Control in Low Light Image and Video Enhancement Using Equivariance

    Sara Aghajanzadeh and David Forsyth. Long scale error con- trol in low light image and video enhancement using equiv- ariance. arXiv:2206.01334, 2022. 3

    work page Pith review arXiv 2022

  8. [5]

    Unsupervised learning for color constancy

    Nikola Bani ´c and Sven Lonˇcari´c. Unsupervised learning for color constancy. In VISIGRAPP, 2017. 7, 8

    2017

Show all 92 references
  1. [6]

    Quasi-unsupervised color constancy

    Simone Bianco and Claudio Cusano. Quasi-unsupervised color constancy. In CVPR, 2019. 3, 5, 7, 8

    2019

  2. [7]

    No-reference image quality assessment in the spatial domain

    Alan Conrad Bovik, Anish Mittal, and Anush Krishna Moor- thy. No-reference image quality assessment in the spatial domain. IEEE TIP, 2012. 4

    2012

  3. [8]

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

    Yuanhao Cai, Hao Bian, Jing Lin, Haoqian Wang, Radu Tim- ofte, and Yulun Zhang. Retinexformer: One-stage retinex- based transformer for low-light image enhancement. In ICCV, 2023. 3, 5

    2023

  4. [9]

    Masactrl: Tuning-free mu- tual self-attention control for consistent image synthesis and editing

    Mingdeng Cao, Xintao Wang, Zhongang Qi, Ying Shan, Xi- aohu Qie, and Yinqiang Zheng. Masactrl: Tuning-free mu- tual self-attention control for consistent image synthesis and editing. In ICCV, 2023. 2, 7

    2023

  5. [10]

    Duygu Ceylan, Chun-Hao Huang, and Niloy J. Mitra. Pix2video: Video editing using image diffusion. In ICCV,

  6. [11]

    An efficient statistical method for image noise level estimation

    Guangyong Chen, Fengyuan Zhu, and Pheng-Ann Heng. An efficient statistical method for image noise level estimation. In ICCV, 2015. 4

    2015

  7. [12]

    Illuminant estimation for color constancy: Why spatial- domain methods work and the role of the color distribution

    Dongliang Cheng, Dilip K Prasad, and Michael S Brown. Illuminant estimation for color constancy: Why spatial- domain methods work and the role of the color distribution. Journal of the Optical Society of America A, 2014. 8

    2014

  8. [13]

    Schnabel, and Tingying Peng

    Tom ´aˇs Chobola, Yu Liu, Hanyi Zhang, Julia A. Schnabel, and Tingying Peng. Fast context-based low-light image en- hancement via neural implicit representations. In ECCV,

  9. [14]

    Ilvr: Conditioning method for denoising diffusion probabilistic models

    Jooyoung Choi, Sungwon Kim, Yonghyun Jeong, Youngjune Gwon, and Sungroh Yoon. Ilvr: Conditioning method for denoising diffusion probabilistic models. In ICCV, 2021. 3

    2021

  10. [15]

    Come- closer-diffuse-faster: Accelerating conditional diffusion models for inverse problems through stochastic contraction

    Hyungjin Chung, Byeongsu Sim, and Jong-Chul Ye. Come- closer-diffuse-faster: Accelerating conditional diffusion models for inverse problems through stochastic contraction. In CVPR, 2021. 3

    2021

  11. [16]

    Style injec- tion in diffusion: A training-free approach for adapting large- scale diffusion models for style transfer

    Jiwoo Chung, Sangeek Hyun, and Jae-Pil Heo. Style injec- tion in diffusion: A training-free approach for adapting large- scale diffusion models for style transfer. In CVPR, 2024. 2

    2024

  12. [17]

    Dhariwal and A

    P. Dhariwal and A. Nichol. Diffusion models beat gans on image synthesis. In NeurIPS, 2021. 3

    2021

  13. [18]

    Diffusion models beat GANs on image synthesis

    Prafulla Dhariwal and Alexander Quinn Nichol. Diffusion models beat GANs on image synthesis. In NeurIPS, 2021. 6

    2021

  14. [19]

    Efros, and Aleksander Holynski

    Dave Epstein, Allan Jabri, Ben Poole, Alexei A. Efros, and Aleksander Holynski. Diffusion self-guidance for control- lable image generation. In NeurIPS, 2023. 2

    2023

  15. [20]

    Generative dif- fusion prior for unified image restoration and enhancement

    Ben Fei, Zhaoyang Lyu, Liang Pan, Junzhe Zhang, Weidong Yang, Tianyue Luo, Bo Zhang, and Bo Dai. Generative dif- fusion prior for unified image restoration and enhancement. In CVPR, 2023. 1, 2, 3, 5

    2023

  16. [21]

    David A. Forsyth. A novel algorithm for color constancy. IJCV, 1990. 1

    1990

  17. [22]

    Learning a simple low-light image enhancer from paired low-light instances

    Zhenqi Fu, Yan Yang, Xiaotong Tu, Yue Huang, Xinghao Ding, and Kai-Kuang Ma. Learning a simple low-light image enhancer from paired low-light instances. In CVPR, 2023. 3, 5

    2023

  18. [23]

    Bayesian color constancy revisited

    Peter Vincent Gehler, Carsten Rother, Andrew Blake, Tom Minka, and Toby Sharp. Bayesian color constancy revisited. In CVPR, 2008. 8

    2008

  19. [24]

    Color in Computer Vision: Fundamentals and Applications

    Theo Gevers, Arjan Gijsenij, Joost Van de Weijer, and Jan- Mark Geusebroek. Color in Computer Vision: Fundamentals and Applications. John Wiley & Sons, 2012. 3

    2012

  20. [25]

    Test-time degradation adaptation for open-set image restoration

    Yuanbiao Gou, Haiyu Zhao, Boyun Li, Xinyan Xiao, and Xi Peng. Test-time degradation adaptation for open-set image restoration. In ICML, 2024. 2, 3, 5, 7, 8

    2024

  21. [26]

    Zero- reference deep curve estimation for low-light image en- hancement

    Chunle Guo Guo, Chongyi Li, Jichang Guo, Chen Change Loy, Junhui Hou, Sam Kwong, and Runmin Cong. Zero- reference deep curve estimation for low-light image en- hancement. In CVPR, 2020. 3, 5

    2020

  22. [27]

    Lime: Low-light im- age enhancement via illumination map estimation

    Xiaojie Guo, Yu Li, and Haibin Ling. Lime: Low-light im- age enhancement via illumination map estimation. IEEE TIP, 2017. 4

    2017

  23. [28]

    R2rnet: Low-light image en- hancement via real-low to real-normal network

    Jiang Hai, Zhu Xuan, Ren Yang, Yutong Hao, Fengzhu Zou, Fang Lin, and Songchen Han. R2rnet: Low-light image en- hancement via real-low to real-normal network. Journal of Visual Communication and Image Representation, 2023. 4

    2023

  24. [29]

    X.J. Shi H.D. Cheng. A simple and effective histogram equalization approach to image enhancement. InDigital Sig- nal Processing, 2004. 2

    2004

  25. [30]

    Style aligned image generation via shared atten- tion

    Amir Hertz, Andrey V oynov, Shlomi Fruchter, and Daniel Cohen-Or. Style aligned image generation via shared atten- tion. In CVPR, 2024. 2

    2024

  26. [31]

    J. Ho, A. Jain, and P. Abbeel. Denoising diffusion proba- bilistic models. In NeurIPS, 2020. 3

    2020

  27. [32]

    Global structure-aware diffusion pro- cess for low-light image enhancement

    Jinhui Hou, Zhiyu Zhu, Junhui Hou, Hui Liu, Huanqiang Zeng, and Hui Yuan. Global structure-aware diffusion pro- cess for low-light image enhancement. In NeurIPS, 2023. 5

    2023

  28. [33]

    Global structure-aware diffusion pro- cess for low-light image enhancement

    Jinhui Hou, Zhiyu Zhu, Junhui Hou, Hui Liu, Huanqiang Zeng, and Hui Yuan. Global structure-aware diffusion pro- cess for low-light image enhancement. In NeurIPS, 2023. 3

    2023

  29. [34]

    Arbitrary style transfer in real-time with adaptive instance normalization

    Xun Huang and Serge Belongie. Arbitrary style transfer in real-time with adaptive instance normalization. In ICCV,

  30. [35]

    Low-light image enhancement with wavelet-based diffusion models

    Hai Jiang, Ao Luo, Haoqiang Fan, Songchen Han, and Shuaicheng Liu. Low-light image enhancement with wavelet-based diffusion models. ACM TOG, 2023. 3

    2023

  31. [36]

    Lightendiffusion: Unsupervised low-light 9 image enhancement with latent-retinex diffusion models

    Hai Jiang, Ao Luo, Xiaohong Liu, Songchen Han, and Shuaicheng Liu. Lightendiffusion: Unsupervised low-light 9 image enhancement with latent-retinex diffusion models. In ECCV, 2024. 3

    2024

  32. [37]

    Enlightengan: Deep light enhancement without paired supervision

    Yifan Jiang, Xinyu Gong, Ding Liu, Yu Cheng, Chen Fang, Xiaohui Shen, Jianchao Yang, Pan Zhou, and Zhangyang Wang. Enlightengan: Deep light enhancement without paired supervision. In CVPR, 2019. 3, 5

    2019

  33. [38]

    Edwin H. Land. The retinex theory of color vision. Scientific American, 1977. 1, 3

    1977

  34. [39]

    Contrast en- hancement based on layered difference representation of 2d histograms

    Chulwoo Lee, Chul Lee, and Chang-Su Kim. Contrast en- hancement based on layered difference representation of 2d histograms. IEEE TIP, 2013. 4

    2013

  35. [40]

    An underwater image enhancement benchmark dataset and beyond

    Chongyi Li, Chunle Guo, Wenqi Ren, Runmin Cong, Junhui Hou, Sam Kwong, and Dacheng Tao. An underwater image enhancement benchmark dataset and beyond. In IEEE TIP,

  36. [41]

    Learning to enhance low-light image via zero-reference deep curve es- timation

    Chongyi Li, Chunle Guo, and Chen Change Loy. Learning to enhance low-light image via zero-reference deep curve es- timation. In IEEE TPAMI, 2021. 3, 5

    2021

  37. [42]

    A deep learning based image enhancement ap- proach for autonomous driving at night

    Guofa Li, Yifan Yang, Xingda Qu, Dongpu Cao, and Ke- qiang Li. A deep learning based image enhancement ap- proach for autonomous driving at night. In Knowledge- Based Systems, 2021. 1

    2021

  38. [43]

    Iterative prompt learning for unsupervised backlit image enhancement

    Zhexin Liang, Chongyi Li, Shangchen Zhou, Ruicheng Feng, and Chen Change Loy. Iterative prompt learning for unsupervised backlit image enhancement. In ICCV, 2023. 3, 5

    2023

  39. [44]

    Barron, Dil- lon Sharlet, Ryan Geiss, Samuel W

    Orly Liba, Kiran Murthy, Yun-Ta Tsai, Tim Brooks, Tian- fan Xue, Nikhil Karnad, Qiurui He, Jonathan T. Barron, Dil- lon Sharlet, Ryan Geiss, Samuel W. Hasinoff, Yael Pritch, and Marc Levoy. Handheld mobile photography in very low light. In ACM TOG, 2019. 1

    2019

  40. [45]

    Belongie, James Hays, Pietro Perona, Deva Ramanan, Piotr Dollar, and C

    Tsung-Yi Lin, Michael Maire, Serge J. Belongie, James Hays, Pietro Perona, Deva Ramanan, Piotr Dollar, and C. Lawrence Zitnick. Microsoft coco: Common objects in context. In ECCV, 2014. 3

    2014

  41. [47]

    Repaint: Inpainting using denoising diffusion probabilistic models

    Andreas Lugmayr, Martin Danelljan, Andr´es Romero, Fisher Yu, Radu Timofte, and Luc Van Gool. Repaint: Inpainting using denoising diffusion probabilistic models. In CVPR,

  42. [48]

    MBLLEN: low-light image/video enhancement using cnns

    Feifan Lv, Feng Lu, Jianhua Wu, and Chongsoon Lim. MBLLEN: low-light image/video enhancement using cnns. In BMVC, 2018. 2

    2018

  43. [49]

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

    Xiaoqian Lv, Shengping Zhang, Chenyang Wang, Yichen Zheng, Bineng Zhong, Chongyi Li, and Liqiang Nie. Fourier priors-guided diffusion for zero-shot joint low-light enhance- ment and deblurring. In CVPR, 2024. 1, 2, 3, 5

    2024

  44. [50]

    Perceptual quality as- sessment for multi-exposure image fusion

    Kede Ma, Kai Zeng, and Zhou Wang. Perceptual quality as- sessment for multi-exposure image fusion. IEEE TIP, 2015. 4

    2015

  45. [51]

    Toward fast, flexible, and robust low-light image enhancement

    Long Ma, Tengyu Ma, Risheng Liu, Xin Fan, and Zhongx- uan Luo. Toward fast, flexible, and robust low-light image enhancement. In CVPR, 2022. 3, 5

    2022

  46. [52]

    Richter, Laura Waller, and Vladlen Koltun

    Kristina Monakhova, Stephan R. Richter, Laura Waller, and Vladlen Koltun. Dancing under the stars: Video denoising in starlight. In CVPR, 2022. 3

    2022

  47. [53]

    Diffusion in the dark: A diffusion model for low-light text recognition

    Cindy M Nguyen, Eric R Chan, Alexander W Bergman, and Gordon Wetzstein. Diffusion in the dark: A diffusion model for low-light text recognition. In WACV, 2024. 3

    2024

  48. [54]

    Zero-shot image-to-image translation

    Gaurav Parmar, Krishna Kumar Singh, Richard Zhang, Yijun Li, Jingwan Lu, and Jun-Yan Zhu. Zero-shot image-to-image translation. In ACM SIGGRAPH, 2023. 2, 7

    2023

  49. [55]

    Object retrieval with large vocabularies and fast spatial matching

    James Philbin, Ondrej Chum, Michael Isard, Josef Sivic, and Andrew Zisserman. Object retrieval with large vocabularies and fast spatial matching. In CVPR, 2007. 8

    2007

  50. [56]

    Pizer, Elton Philip Amburn, John D

    Stephen M. Pizer, Elton Philip Amburn, John D. Austin, Robert Cromartie, Ari Geselowitz, Trey Greer, Bart M. ter Haar Romeny, and John B. Zimmerman. Adaptive histogram equalization and its variations. Graphical Models graphical Models and Image Processing computer Vision, Grap...

    1987

  51. [57]

    SDXL: Improving latent diffusion models for high-resolution image synthesis

    Dustin Podell, Zion English, Kyle Lacey, Andreas Blattmann, Tim Dockhorn, Jonas M ¨uller, Joe Penna, and Robin Rombach. SDXL: Improving latent diffusion models for high-resolution image synthesis. In ICLR, 2024. 3

    2024

  52. [58]

    L2uwe: A framework for the efficient enhancement of low-light un- derwater images using local contrast and multi-scale fusion

    Tunai Porto Marques and Alexandra Branzan Albu. L2uwe: A framework for the efficient enhancement of low-light un- derwater images using local contrast and multi-scale fusion. In CVPRW, 2020. 1

    2020

  53. [59]

    Fatezero: Fus- ing attentions for zero-shot text-based video editing

    Chenyang Qi, Xiaodong Cun, Yong Zhang, Chenyang Lei, Xintao Wang, Ying Shan, and Qifeng Chen. Fatezero: Fus- ing attentions for zero-shot text-based video editing. ICCV,

  54. [60]

    Rahman, M

    S. Rahman, M. M. Rahman, M. Abdullah-Al-Wadud, G. D. Al-Quaderi, and M. Shoyaib. An adaptive gamma correction for image enhancement. EURASIP Journal on Image and Video Processing, 2016. 2

    2016

  55. [61]

    Retinex-inspired unrolling with cooperative prior architecture search for low-light image enhancement

    Liu Risheng, Ma Long, Zhang Jiaao, Fan Xin, and Luo Zhongxuan. Retinex-inspired unrolling with cooperative prior architecture search for low-light image enhancement. In CVPR, 2021. 3, 5

    2021

  56. [62]

    Rombach, A

    R. Rombach, A. Blattmann, D. Lorenz, P. Esser, and B. Om- mer. High-resolution image synthesis with latent diffusion models. In CVPR, 2022. 3

    2022

  57. [63]

    High-resolution image syn- thesis with latent diffusion models

    Robin Rombach, Andreas Blattmann, Dominik Lorenz, Patrick Esser, and Bjorn Ommer. High-resolution image syn- thesis with latent diffusion models. In CVPR, 2022. 8

    2022

  58. [64]

    Berg, and Li Fei-Fei

    Olga Russakovsky, Jia Deng, Hao Su, Jonathan Krause, San- jeev Satheesh, Sean Ma, Zhiheng Huang, Andrej Karpathy, Aditya Khosla, Michael Bernstein, Alexander C. Berg, and Li Fei-Fei. Imagenet large scale visual recognition challenge. IJCV, 2015. 8

    2015

  59. [65]

    Lit the darkness: Three- stage zero-shot learning for low-light enhancement with multi-neighbor enhancement factors

    Mariam Saeed and Marwan Torki. Lit the darkness: Three- stage zero-shot learning for low-light enhancement with multi-neighbor enhancement factors. In ICASSP, 2023. 3, 5

    2023

  60. [66]

    Zero-ig: Zero-shot illumination-guided joint denoising and adaptive enhancement for low-light images

    Yiqi Shi, Duo Liu, Liguo Zhang, Ye Tian, Xuezhi Xia, and Xiaojing Fu. Zero-ig: Zero-shot illumination-guided joint denoising and adaptive enhancement for low-light images. In CVPR, 2024. 3, 5

    2024

  61. [67]

    Dragdiffusion: Harnessing diffusion models for interactive point-based image editing

    Yujun Shi, Chuhui Xue, Jiachun Pan, Wenqing Zhang, Vin- cent YF Tan, and Song Bai. Dragdiffusion: Harnessing diffusion models for interactive point-based image editing. CVPR, 2024. 2, 7 10

    2024

  62. [68]

    Sohl-Dickstein, E

    J. Sohl-Dickstein, E. Weiss, N. Maheswaranathan, and S. Ganguli. Deep unsupervised learning using nonequilibrium thermodynamics. In ICML, 2015. 3

    2015

  63. [69]

    Denois- ing diffusion implicit models

    Jiaming Song, Chenlin Meng, and Stefano Ermon. Denois- ing diffusion implicit models. In ICLR, 2021. 4, 6, 8

    2021

  64. [70]

    Song and S

    Y . Song and S. Ermon. Generative modeling by estimating gradients of the data distribution. In NeurIPS, 2019. 3

    2019

  65. [71]

    Y . Song, J. Sohl-Dickstein, D. P. Kingma, A. Kumar, S. Ermon, and B. Poole. Score-based generative modeling through stochastic differential equations. In ICLR, 2021. 3

    2021

  66. [72]

    Plug-and-play diffusion features for text-driven image-to-image translation

    Narek Tumanyan, Michal Geyer, Shai Bagon, and Tali Dekel. Plug-and-play diffusion features for text-driven image-to-image translation. In CVPR, 2023. 2, 7

    2023

  67. [73]

    Timothy Alexis Vass, 2024. 3

    2024

  68. [74]

    On the evaluation of illumination compensation algorithms

    Vassilios V onikakis, Rigas Kouskouridas, and Antonios Gasteratos. On the evaluation of illumination compensation algorithms. In Multimedia Tools and Applications, 2018. 4

    2018

  69. [75]

    Nat- uralness preserved enhancement algorithm for non-uniform illumination images

    Shuhang Wang, Jin Zheng, Hai-Miao Hu, and Bo Li. Nat- uralness preserved enhancement algorithm for non-uniform illumination images. IEEE TIP, 2013. 4

    2013

  70. [77]

    Zero-shot image restoration using denoising diffusion null-space model

    Yinhuai Wang, Jiwen Yu, and Jian Zhang. Zero-shot image restoration using denoising diffusion null-space model. In ICLR, 2023. 3, 5, 7, 8

    2023

  71. [78]

    Kot, and Bihan Wen

    Yufei Wang, Yi Yu, Wenhan Yang, Lanqing Guo, Lap-Pui Chau, Alex C. Kot, and Bihan Wen. Exposurediffusion: Learning to expose for low-light image enhancement. In ICCV, 2023. 3

    2023

  72. [79]

    Deep retinex decomposition for low-light enhancement

    Chen Wei, Wenjing Wang, Wenhan Yang, and Jiaying Liu. Deep retinex decomposition for low-light enhancement. In BMVC, 2018. 2, 4, 8

    2018

  73. [81]

    Coarse-to-fine luminance estimation for low-light image enhancement in maritime video surveillance

    Meifang Yang, Xin Nie, and Ryan Wen Liu. Coarse-to-fine luminance estimation for low-light image enhancement in maritime video surveillance. In IEEE ITSC, 2019. 2

    2019

  74. [82]

    Implicit neural representation for cooperative low-light image enhancement

    Shuzhou Yang, Moxuan Ding, Yanmin Wu, Zihan Li, and Jian Zhang. Implicit neural representation for cooperative low-light image enhancement. In ICCV, 2023. 3, 5

    2023

  75. [83]

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

    Wenhan Yang, Wenjing Wang, Haofeng Huang, Shiqi Wang, and Jiaying Liu. Sparse gradient regularized deep retinex network for robust low-light image enhancement. IEEE TIP,

  76. [84]

    X. Yi, H. Xu, H. Zhang, L. Tang, and J. Ma. Diff-retinex: Rethinking low-light image enhancement with a generative diffusion model. In ICCV, 2023. 3

    2023

  77. [85]

    A feature-enriched completely blind image quality evaluator

    Lin Zhang, Lei Zhang, and Alan C Bovik. A feature-enriched completely blind image quality evaluator. IEEE TIP, 2015. 4

    2015

  78. [86]

    Zero-shot restoration of back-lit images using deep internal learning

    Lin Zhang, Lijun Zhang, Xinyu Liu, Ying Shen, Shaoming Zhang, and Shengjie Zhao. Zero-shot restoration of back-lit images using deep internal learning. In ACM MM, 2019. 2, 3, 5

    2019

  79. [87]

    Efros, Eli Shecht- man, and Oliver Wang

    Richard Zhang, Phillip Isola, Alexei A. Efros, Eli Shecht- man, and Oliver Wang. The unreasonable effectiveness of deep features as a perceptual metric. In CVPR, 2018. 4

    2018

  80. [88]

    Kindling the darkness: A practical low-light image enhancer

    Yonghua Zhang, Jiawan Zhang, and Xiaojie Guo. Kindling the darkness: A practical low-light image enhancer. In ACM MM, 2019. 5

    2019

  81. [89]

    Beyond brightening low-light images

    Yonghua Zhang, Xiaojie Guo, Jiayi Ma, Wei Liu, and Jiawan Zhang. Beyond brightening low-light images. IJCV, 2021. 5

    2021

  82. [90]

    Semantic-guided zero-shot learning for low-light image/video enhancement

    Shen Zheng and Gaurav Gupta. Semantic-guided zero-shot learning for low-light image/video enhancement. In WACV,

  83. [91]

    Places: A 10 million image database for scene recognition

    Bolei Zhou, Agata Lapedriza, Aditya Khosla, Aude Oliva, and Antonio Torralba. Places: A 10 million image database for scene recognition. IEEE TPAMI, 2017. 8

    2017

  84. [92]

    Zero-shot restoration of underexposed im- ages via robust retinex decomposition

    Anqi Zhu, Lin Zhang, Ying Shen, Yong Ma, Shengjie Zhao, and Yicong Zhou. Zero-shot restoration of underexposed im- ages via robust retinex decomposition. In ICME, 2020. 3, 5 11

    2020

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.