Pith. sign in

REVIEW 5 major objections 5 minor 2 cited by

Learning Adaptive Lighting via Channel-Aware Guidance

T0 review · 5 major / 5 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read A single network that tracks each color channel's light response separately outperforms task-specific methods on four light-related image enhancement benchmarks while using fewer computational resources.

desk verdict A serious, well-engineered enhancement architecture with a novel color-separated guidance design and strong benchmark claims, held back mostly by missing code, unspecified loss weights, and internal number inconsistencies. read the letter →

arxiv 2412.01493 v2 pith:PSHHVUF4 submitted 2024-12-02 cs.CV eess.IV

Qirui Yang , Peng-Tao Jiang , Hao Zhang , Jinwei Chen , Bo Li , Huanjing Yue , Jingyu Yang This is my paper · ORCID
classification cs.CVeess.IV
keywords lightadaptationmulti-taskimageenhancementchannel-awareattentionexposurecorrectiontonemappinglow-lightretouchingstatespacemodel
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 claims that exposure correction, image retouching, tone mapping, and low-light enhancement share a common structure: the red, green, and blue channels carry different light information, and these channel differences show up differently in the spatial domain than in the frequency domain. Building on that, LALNet keeps RGB channels separate in one branch while mixing them in another, and lets the separated features guide the mixed features through an attention mechanism. The payoff, if the claim is right, is that one architecture beats methods designed for each individual task and does it with less computation. On the reported benchmarks the gains are 0.86 dB PSNR on SCIE, 2.15 dB on HDR+, 1.23 dB on LOLv1, and 3.86 dB on HDRI Haven over the second-best method.

What carries the argument

The load-bearing mechanism is Light Guided Attention (LGA), an attention module in which the query comes from color-separated features (each RGB channel processed independently) and the keys and values come from color-mixed features, so the network is forced to attend to channel differences while integrating global lighting. Supporting it are Dual Domain Channel Modulation (DDCM), which applies learnable convolutions to the real and imaginary parts of each channel's FFT and then channel attention in the spatial domain; Mixed Channel Modulation (MCM) with wavelet transform; and the Light State Space Module (LSSM), a selective-scan state space module that captures global context at low cost. A learnable differential pyramid plus iterative detail enhancement restores high frequencies after light adaptation is done at low resolution.

What would settle it

Compute per-channel luminance means and DWT or FFT energy distributions over the full training and test splits of SCIE, HDR+, LOLv1, and HDRI Haven; if the channel-difference patterns do not systematically vary between tasks, or if replacing LGA's channel-specific queries with a shared query leaves the reported PSNR gains unchanged, the central premise fails.

Watch

Extended reading notes

Core claim

LALNet's central discovery is that the channel-wise light differences are the shared currency across four light-related tasks, and that a network can exploit them explicitly. The design separates color information: a Dual Domain Channel Modulation branch processes each RGB channel alone in both Fourier and spatial domains to produce color-separated features, while a mixed branch with wavelet modulation and a state space module produces color-mixed features. Light Guided Attention then injects the separated features into the mixed stream as the query, forcing the network to keep channel differences in view while it adapts global lighting. The paper reports that this design outperforms all compared methods on four benchmarks and that the main model uses 6.70 GFLOPs on LOLv1 where the prior best uses 42.82 GFLOPs.

Load-bearing premise

The whole architecture assumes that color channels carry reliably different light information in both spatial and frequency domains, and that this difference is shared enough across the four tasks to be worth a dedicated branch; the paper illustrates this with only two example images and no dataset-wide statistics.

Editorial extensions

If this is right

  • One trained architecture can be adapted to exposure correction, retouching, low-light enhancement, and tone mapping without redesigning the network, since only the training data changes.
  • Color-separated guidance is what carries the reported gains: removing LGA drops PSNR by 1.11 dB on SCIE, and replacing group convolution with ordinary convolution drops 0.64 dB.
  • Efficiency follows from doing light adaptation at low resolution: LALNet uses 6.70 GFLOPs on LOLv1 versus 42.82 for RetinexMamba while still improving PSNR by 1.23 dB.
  • The lightweight variant LALNet-Tiny, with 230K parameters and 1.75 GFLOPs, beats the best prior method on HDR+ and LOLv1, suggesting the design scales down without collapsing.
  • Because the same framework tops benchmarks on four tasks, the shared channel-difference property is a workable design principle for multi-task lighting models.

Reading between the lines

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

  • If the channel-difference statistics are stable within a task, the same LGA mechanism could be applied to video tone mapping or RAW sensor data, where per-channel noise profiles differ even more strongly than in sRGB.
  • A direct test of the paper's premise would be to measure per-channel luminance and spectral-energy differences across the full SCIE, HDR+, LOL, and HDRI Haven training sets; the paper shows only two examples, so the claimed generality is not yet statistically demonstrated.
  • The success of using FFT-domain per-channel convolutions hints that a physical image-formation model, such as illuminant color times sensor response, could supply the channel priors instead of learned modulation, making the mechanism more inspectable.
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

5 major / 5 minor

Summary. The paper proposes LALNet, a multi-task network for four light-related image enhancement tasks: exposure correction, image retouching, low-light enhancement, and tone mapping. The architecture uses a dual-branch design: a color-separated branch (DDCM and group convolution) that processes RGB channels independently in the spatial and frequency domains, and a color-mixed branch (MCM and LSSM) that captures cross-channel and global lighting information. Light Guided Attention (LGA) injects color-separated features into color-mixed features. Reconstruction is performed with a learnable differential pyramid and iterative detail enhancement. The authors report state-of-the-art PSNR/SSIM results on SCIE, HDR+, LOLv1, and HDRI Haven, along with lower GFLOPs for the full model and a lightweight LALNet-Tiny variant.

Significance. If the reported results are reproducible, LALNet would be a useful contribution: it is a single architecture applicable to four light-related tasks, with a clear design story based on channel-separated versus channel-mixed features, strong quantitative gains on standard benchmarks, and a lightweight variant (230K parameters, 1.75 GFLOPs). The paper also includes ablation studies on the key modules and third-party generalization tests in the appendix, which are valuable. However, the significance is currently conditional because key training details (loss weights, values of tau and eta_C) are missing, the ablation table is unreadable, and no code or weights are provided; the reader cannot currently verify the central claim. The wording 'significantly outperforms' is also not backed by any statistical test.

major comments (5)
  1. [Appendix E, Eq. (19); Sec. 4.1] The loss weights alpha, beta, gamma, and eta in Eq. (19) are never specified. Section 4.1 reports the optimizer, learning rate, and iteration count but omits the coefficients of the four-term objective. Eq. (16) also leaves tau as an unspecified scaling factor, and the LSSM expansion factor eta_C introduced after Eq. (12) is not given. Because the loss balance and these hyperparameters materially affect the reported PSNR/SSIM numbers, the claimed gains cannot be independently reproduced or checked. Please provide the loss coefficients, tau, eta_C, and the code or pretrained weights; the current online demo link is a placeholder.
  2. [Sec. 5 vs. Table 4; Sec. 4.2] The conclusion states a 1.12 dB PSNR gain on LOL, but Table 4 shows 25.26 versus 24.03 dB, i.e., 1.23 dB, and Section 4.2 also says 1.23 dB. In addition, no standard deviations, confidence intervals, or significance tests are reported for any of the benchmark tables, so the abstract's 'significantly outperforms' is not statistically supported. Please correct the numerical inconsistency and add variance estimates or significance tests, or temper the language.
  3. [Sec. 4.3, Table 5] The ablation matrix is unreadable. The text describes check/cross markers for MCM, DDCM, LGA, and LSSM, but Table 5 prints '%' and '!' symbols, so the reader cannot determine which module is removed in each variant row. The description in the text (e.g., #2 replaces MCM, #3 replaces DDCM, #4 removes LGA, #5 replaces LSSM) must be encoded unambiguously; as printed, the table does not support the claimed per-module contributions.
  4. [Sec. 3.4 and Sec. 4.2 (HDRI Haven)] LALNet reuses the learnable differential pyramid (LDP) from the authors' concurrent paper (Yang et al. 2024, arXiv:2412.01463) and reports a 3.86 dB PSNR gain over the next-best method on the HDRI Haven tone-mapping benchmark. Because LDP is a strong component taken from another paper and no ablation is provided that replaces LDP with a standard upsampling or pyramid scheme, the marginal contribution of the proposed channel-aware modules on this benchmark is unclear. Please clarify the evaluation protocol on HDRI Haven--including the origin of the train/test split and whether baseline numbers are reproduced under the same settings--and add an ablation that removes the LDP component or reruns the closest baseline under identical conditions.
  5. [Sec. 4.2 and abstract] The claim that LALNet 'requires fewer computational resources' is only partially quantified. Table 4 reports GFLOPs for the LOL benchmark, and the text gives 1.75 GFLOPs for LALNet-Tiny, but Tables 1-3 do not report GFLOPs for LALNet on SCIE, HDR+, or HDRI Haven. The cross-task efficiency claim should be supported by consistent FLOPs/GFLOPs reporting across all four benchmarks.
minor comments (5)
  1. [Fig. 4, Fig. 5, Fig. 9, Fig. 10 captions] Several captions and the surrounding text use 'LALet' instead of 'LALNet'; please correct the typo.
  2. [Table 7 caption] The caption says 'MSCE' while the text and table header say 'MSEC'; the dataset name should be consistent.
  3. [Abstract] The online demo URL (https://xxxxxx2025.github.io/LALNet/) is a placeholder and should be replaced with a working link or removed.
  4. [Eq. (16) and Sec. 4.1] The scaling factor tau in Eq. (16) is not given a value or a tuning schedule; if it is a fixed hyperparameter, please report it in the implementation details.
  5. [Appendix B] The repeated IDE formulation uses inconsistent subscript notation (Y_i versus Y_l); please align the notation with Eq. (17).

Circularity Check

0 steps flagged · score 1.0 of 10

No material circularity: the central claim rests on external benchmark comparisons; only minor self-citation overlap appears.

full rationale

The paper's derivation chain is not circular. The central claim that LALNet outperforms prior methods on four light-related tasks is supported by quantitative comparisons in Tables 1-4 against external published methods on standard benchmarks, including SCIE, HDR+, LOLv1, and HDRI Haven. None of the reported PSNR/SSIM values is defined in terms of the paper's channel-difference premise: DDCM, MCM, LSSM, and LGA are trainable modules optimized by a loss (Eq. 19) whose target is the reference image, not a summary of channel statistics. The motivational analysis in Sec. 1 and Fig. 2 is illustrative rather than a fitted input; no parameter is fit to the benchmark outputs and then renamed a prediction. The only self-citation that enters the architecture is the learnable differential pyramid (Yang et al., 2024), and the HDRI Haven benchmark also originates from that prior work; however, these are component and dataset choices, not the multi-task channel-guidance claim, and the comparisons include many external methods. The absence of code/weights, the unspecified loss weights in Appendix E, and the ambiguous ablation markers in Table 5 are reproducibility and correctness risks, not circularity. Score 1 reflects the minor self-citation overlap with no load-bearing circular step.

Assumptions & free parameters 5 free parameters · 6 assumptions · 0 invented entities

The central empirical load is carried by benchmark comparisons, not derived laws. The architectural motivation depends on qualitative channel-difference observations from two illustrations, and the efficiency claim depends on the assumption that a heavily downsampled low-frequency input retains the information needed for light adaptation. Several hyperparameters that affect results are unreported. No new physical entities are introduced.

free parameters (5)
  • Loss weights alpha, beta, gamma, eta (Eq. 19) = Not reported
    The combined loss in Eq. 19 mixes reconstruction, SSIM, perceptual, and high-frequency terms; the weights are never given, so the exact training objective is underspecified.
  • LGA scaling factor tau (Eq. 16) = Not reported
    Temperature-like scale in the attention softmax; value omitted.
  • LSSM expansion factor eta_C = Not reported
    The first stream in LSSM expands features to eta times C channels; eta is not specified.
  • Pyramid levels L = 3
    Ablation Table 6 selects L=3; L=4 lowers PSNR to 23.07 and L=1 exceeds GPU memory, so the choice is data-driven.
  • LALNet-Tiny widths/depths = Not reported
    The Tiny variant is described only as fewer channels and fewer LSSM blocks; exact widths are absent.
assumptions (6)
  • domain assumption Color channels exhibit distinct light properties, and these differences differ between spatial and frequency domains (Sec. 1, Fig. 2).
    Used to motivate DDCM and LGA; only two example image pairs are shown, with no statistical test across datasets.
  • domain assumption These channel-level properties are shared across exposure correction, retouching, tone mapping, and low-light enhancement (Sec. 3.1).
    Basis for the multi-task architecture claim; no cross-task transfer experiment with a single shared model is reported.
  • domain assumption A heavily downsampled low-frequency image retains enough light information, and lost details can be recovered by the iterative pyramid (Sec. 3.2, Eqs. 17-18).
    Supports the low-compute claim; only validated indirectly via pyramid-level ablation on the SCIE dataset.
  • domain assumption The learnable differential pyramid of Yang et al. 2024 is accepted as a building block.
    Adopted from the authors' own prior work without re-derivation in this paper.
  • standard math Standard FFT/IFFT and discrete wavelet transforms are invertible and preserve the needed information (Eqs. 4-7, Eqs. 9-10).
    Standard transforms; not at issue.
  • domain assumption Benchmark metrics (PSNR, SSIM, LPIPS, TMQI, CIELAB Delta E) are valid proxies for perceptual quality in these tasks (Sec. 4.1).
    Performance claims are defined in terms of these metrics.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Learning Adaptive Lighting via Channel-Aware Guidance." pith.science (2026). https://pith.science/paper/PSHHVUF4

@misc{pith2026241201493,
  author       = {Pith},
  title        = {Pith review of: Learning Adaptive Lighting via Channel-Aware Guidance},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/PSHHVUF4}},
  note         = {Machine review of arXiv:2412.01493}
}
read the original abstract

Learning lighting adaptation is a crucial step in achieving good visual perception and supporting downstream vision tasks. Current research often addresses individual light-related challenges, such as high dynamic range imaging and exposure correction, in isolation. However, we identify shared fundamental properties across these tasks: i) different color channels have different light properties, and ii) the channel differences reflected in the spatial and frequency domains are different. Leveraging these insights, we introduce the channel-aware Learning Adaptive Lighting Network (LALNet), a multi-task framework designed to handle multiple light-related tasks efficiently. Specifically, LALNet incorporates color-separated features that highlight the unique light properties of each color channel, integrated with traditional color-mixed features by Light Guided Attention (LGA). The LGA utilizes color-separated features to guide color-mixed features focusing on channel differences and ensuring visual consistency across all channels. Additionally, LALNet employs dual domain channel modulation for generating color-separated features and a mixed channel modulation and light state space module for producing color-mixed features. Extensive experiments on four representative light-related tasks demonstrate that LALNet significantly outperforms state-of-the-art methods on benchmark tests and requires fewer computational resources. We provide an anonymous online demo at https://xxxxxx2025.github.io/LALNet/.

Figures

Figures reproduced from arXiv: 2412.01493 by the authors.

Figure 1
Figure 1. Our LALNet significantly outperforms state-of-the-art methods on four representative benchmark tests of light-related image enhancement, including image retouching, tone mapping, low-light enhancement, and exposure correction. images taken under poor lighting conditions often exhibit degraded quality, which not only affects visual presentation but also poses challenges for tasks such as object detection and tracking… view at source ↗
Figure 2
Figure 2. Motivation of LALNet. Different color channel differences and statistical DWT spectral energy distributions for different tasks. example, exposure correction (Huang et al., 2022b; Zhang et al., 2019b) must adjust the brightness of both underex￾posed and overexposed scenes to achieve clearer images; image retouching (Wang et al., 2023; Su et al., 2024) aims to enhance the aesthetic visual quality of images affected b… view at source ↗
Figure 3
Figure 3. Architecture of LALNet for light adaptation. The core modules of LALNet are: (a) dual domain channel modulation (DDCM) that extracts color-separated features, focusing on light differences for each channel in the spatial and frequency domains, and (b) light guided attention (LGA) utilizes color-separated features to guide color-mixed light information for light adaptation. stance, Retinexformer focuses on separating… view at source ↗
Figures from the paper (7 more)
Figure 4
Figure 4. Figure 4: Visual comparisons between our LALet and the SOTA methods on the SCIE dataset. (Zoom in for best view.) The error maps in the upper left corner facilitate a more precise determination of performance differences. performance differences of various methods, we present an…
Figure 5
Figure 5. Figure 5: Visual comparisons between our LALet and the state-of-the-art methods on the HDRI Haven dataset (Zoom-in for best view). The error maps in the upper left corner facilitate a more precise determination of performance differences [PITH_FULL_IMAGE:figures/full_fig_p009_5.png]
Figure 6
Figure 6. Figure 6: Motivation. Visualization of the light-related task images in different color channels and their corresponding DWT spectra energy distribution. R-FFT denotes the Fourier Frequency Domain diagram of the R channel. LowFreq and HighFreq are low-frequency and high-frequenc…
Figure 7
Figure 7. Figure 7: The architecture of the Iterative Detail Enhancement module progressively restores resolution and fine details. LinearConv silu silu SS2D 𝐹𝐹𝑐𝑐𝑐𝑐 𝑖𝑖 𝐹𝐹𝑙𝑙𝑙𝑙 𝑖𝑖−1 Linear 𝐹𝐹𝑐𝑐𝑐𝑐 𝑖𝑖 SS2D [PITH_FULL_IMAGE:figures/full_fig_p015_7.png]
Figure 8
Figure 8. Figure 8: The architecture of the Light State Space Module. C. More Results We further evaluate the effectiveness of our model on the exposure correction (Afifi et al., 2021), HDR Survey (Fairchild, 2023), and UVTM (Cao et al., 2023) datasets, all of which present more complex l…
Figure 9
Figure 9. Figure 9: Visual comparisons between our LALet and the state-of-the-art methods on the HDR+ dataset. D. Ablation Study To validate the effectiveness of the SS2D module, we use Self-Attention and Residual Block to replace the SS2D module in the original published model. We use th…
Figure 10
Figure 10. Figure 10: Visual comparisons between our LALet and the SOTA methods on the LOLv1 dataset. 17 [PITH_FULL_IMAGE:figures/full_fig_p017_10.png]

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. LLaPa: A Vision-Language Model Framework for Counterfactual-Aware Procedural Planning

    cs.CL 2025-07 conditional novelty 6.0 of 10

    A VLM-based planner with task-oriented segmentation reranking and a clause-level condition retriever reports state-of-the-art scores on ActPlan-1K and ALFRED, though the reported ablation numbers are internally inconsistent.

  2. CWNet: Causal Wavelet Network for Low-Light Image Enhancement

    cs.CV 2025-07 conditional novelty 4.0 of 10

    CWNet mixes wavelet-based frequency enhancement, Mamba-style high-frequency scanning, and two semantic consistency losses to produce competitive low-light image enhancement with 1.23 million parameters.

Reference graph

Works this paper leans on

18 extracted references · 12 canonical work pages · cited by 2 Pith papers

  1. [6]

    Deep Feature Consistent Deep Image Transformations: Downscaling, Decolorization and HDR Tone Mapping

    Hou, X., Duan, J., and Qiu, G. Deep feature consistent deep image transformations: Downscaling, decolorization and hdr tone mapping.arXiv preprint arXiv:1707.09482,

    work page Pith review arXiv

  2. [8]

    To compensate for the loss of details, we use an iterative detail enhancement module to recover high-frequency details

    To reduce the computational resources, we implement light adaptation at low resolution. To compensate for the loss of details, we use an iterative detail enhancement module to recover high-frequency details. Specifically, as shown in Fig. 7, we first up-sample the low-frequency mapped image Yi LF and concatenate it with the HF component Xi−1 HF , then fee...

    work page 2021

  3. [9]

    J., and Kim, C.-S

    Kim, H.-U., Koh, Y . J., and Kim, C.-S. Global and local enhancement networks for paired and unpaired image enhancement. InComputer Vision–ECCV 2020: 16th European Conference, Glasgow, UK, August 23–28, 2020, Proceedings, Part XXV 16, pp. 339–354. Springer,

    work page 2020

  4. [10]

    URL http://dx.doi

    2352/cic.2007.15.1.art00044. URL http://dx.doi. org/10.2352/cic.2007.15.1.art00044. Finder, S. E., Amoyal, R., Treister, E., and Freifeld, O. Wavelet convolutions for large receptive fields.arXiv preprint arXiv:2407.05848,

    arXiv 2007

  5. [11]

    Li, C., Guo, C., Ai, Q., Zhou, S., and Loy, C. C. Flexi- ble piecewise curves estimation for photo enhancement. arXiv preprint arXiv:2010.13412,

    arXiv 2010

  6. [12]

    W., and Kautz, J

    Paris, S., Hasinoff, S. W., and Kautz, J. Local laplacian filters. InACM SIGGRAPH 2011 papers, Jul

    work page 2011

  7. [14]

    Learning to See Low-Light Images via Feature Domain Adaptation

    Yang, K.-F., Cheng, C., Zhao, S.-X., Yan, H.-M., Zhang, X.- S., and Li, Y .-J. Learning to adapt to light.International Journal of Computer Vision, 131(4):1022–1041, 2023a. Yang, Q., Yue, H., Zhang, L., Liu, Y ., Yang, J., et al. Learn- ing to see low-light images via feature domain adaptation. arXiv preprint arXiv:2312.06723, 2023b. Yang, Q., Li, Y ., Ji...

    work page Pith review arXiv

  8. [15]

    Sparse gradient regularized deep retinex network for ro- bust low-light image enhancement.IEEE Transactions on Image Processing, 30:2072–2086,

    Yang, W., Wang, W., Huang, H., Wang, S., and Liu, J. Sparse gradient regularized deep retinex network for ro- bust low-light image enhancement.IEEE Transactions on Image Processing, 30:2072–2086,

    work page

Show all 18 references
  1. [16]

    W., Arora, A., Khan, S., Hayat, M., Khan, F

    12 Learning Adaptive Lighting via Channel-Aware Guidance Zamir, S. W., Arora, A., Khan, S., Hayat, M., Khan, F. S., Yang, M.-H., and Shao, L. Learning enriched features for real image restoration and enhancement. InComputer Vision–ECCV 2020: 16th European Conference, Glasgow, ...

    2020

  2. [18]

    Likewise, the UVTM dataset contains 20 real-world HDR videos, also without ground truth

    The HDR Survey dataset comprises 105 HDR images and is widely adopted for benchmarking HDR tone mapping methods (Cao et al., 2020; Rana et al., 2020; Panetta et al., 2021; Liang et al., 2018; Paris et al., 2011), though it does not provide ground-truth references. Likewise, th...

    2020

  3. [2009]

    doi: 10.1016/j.neuron.2009. 11.028. URL http://dx.doi.org/10.1016/j. neuron.2009.11.028. Shensa, M. J. et al. The discrete wavelet transform: wedding the a trous and mallat algorithms.IEEE Transactions on signal processing, 40(10):2464–2482,

    2009 doi

  4. [2016]

    Conditional sequen- tial modulation for efficient global image retouching

    He, J., Liu, Y ., Qiao, Y ., and Dong, C. Conditional sequen- tial modulation for efficient global image retouching. In Computer Vision–ECCV 2020: 16th European Confer- ence, Glasgow, UK, August 23–28, 2020, Proceedings, Part XIII 16, pp. 679–695. Springer,

    2020

  5. [2017]

    Hu, L., Chen, H., and Allebach, J. P. Joint multi-scale tone mapping and denoising for hdr image enhancement. In 2022 IEEE/CVF Winter Conference on Applications of Computer Vision Workshops (WACVW), Jan

    2022

  6. [2020]

    Mambair: A simple baseline for image restoration with state-space model.arXiv preprint arXiv:2402.15648,

    Guo, H., Li, J., Dai, T., Ouyang, Z., Ren, X., and Xia, S.-T. Mambair: A simple baseline for image restoration with state-space model.arXiv preprint arXiv:2402.15648,

    arXiv

  7. [2021]

    Retinexmamba: Retinex-based mamba for low-light image enhancement.arXiv preprint arXiv:2405.03349,

    Bai, J., Yin, Y ., and He, Q. Retinexmamba: Retinex-based mamba for low-light image enhancement.arXiv preprint arXiv:2405.03349,

    arXiv

  8. [2022]

    Deep retinex decomposition for low-light enhancement.arXiv preprint arXiv:1808.04560,

    Wei, C., Wang, W., Yang, W., and Liu, J. Deep retinex decomposition for low-light enhancement.arXiv preprint arXiv:1808.04560,

    arXiv

  9. [2023]

    Jiang, Y ., Gong, X., Liu, D., Cheng, Y ., Fang, C., Shen, X., Yang, J., Zhou, P., and Enlightengan, Z. W. Deep light enhancement without paired supervision., 2021, 30.DOI: https://doi. org/10.1109/TIP, pp. 2340–2349,

    2021 doi

  10. [2024]

    Learning a deep single image contrast enhancer from multi-exposure images

    Cai, J., Gu, S., and Zhang, L. Learning a deep single image contrast enhancer from multi-exposure images. IEEE Transactions on Image Processing, 27(4):2049– 2062,

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.