Pith. sign in

REVIEW 4 major objections 4 minor 40 references

Retrieval-Augmented Generation-Based Color Restoration for Low-Light Image Enhancement

T0 review · 4 major / 4 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 This paper establishes that residual color bias in low-light image enhancement can be corrected by a modular post-processing stage that retrieves a reference image from an external knowledge base and injects its global color distribution…

desk verdict A genuine, honestly-reported modular color-restoration post-processor for LLIE, but the abstract overclaims consistency against its own Table 3 and the missing baselines/error bars need attention. read the letter →

arxiv 2608.08211 v1 pith:I67LXAKH submitted 2026-08-08 eess.IV cs.CV

Li-Wei Lu , Shaou-Gang Miaou This is my paper · ORCID
classification eess.IVcs.CV
keywords low-lightimageenhancementcolorrestorationretrieval-augmentedgenerationadaptiveinstancenormalizationreference-basedcolorizationFAISSretrievalcastcorrection
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 the color casts remaining in low-light image enhancement—greenish skies, yellow faces, warm whites—persist because end-to-end training couples brightness, structure, and color in one network, leaving color weakly supervised. The authors propose a modular post-processing stage, built on retrieval-augmented generation, that retrieves a similar reference image from an external knowledge base and injects that reference's global color distribution into a small network that predicts a color residual over the front end's output. Across LOLv1, LOLv2-Real, and LOLv2-Synthetic, the module lowers color-specific errors such as ΔE2000 and MAEab for several different front ends without retraining them. The paper's central claim is that color restoration can be decoupled from brightness enhancement and supplied by an external, per-image color prior rather than by learned parametric priors alone.

What carries the argument

The mechanism is a three-part pipeline: a dual-index FAISS retriever that represents each image by the global mean (texture) and global variance (structure) of its VGG19 relu3_4 features and re-ranks candidates with equal weights; GlobalSPHistAdaIN, which converts the reference's spatial-preserving color histogram into a 512-dimensional global color vector and modulates encoder features by instance normalization with per-channel scale and shift generated from that vector; and a residual formulation in which the network outputs Δab added to the front end's ab channels. The design removes pixel-level correspondence between reference and target, which the authors argue is essential because a retrieved image need not be spatially aligned with the scene.

What would settle it

One concrete experiment: take a fixed set of test images, replace the Top-1 retrieved reference with a deliberately color-shifted reference (e.g., raised a-channel and lowered b-channel), and measure whether color metrics degrade; the paper's mechanism predicts they should, since the injection is driven entirely by global statistics, whereas a retrieval-free network should be unaffected.

Watch

Extended reading notes

Core claim

The central claim is that color restoration in low-light enhancement should be decoupled from brightness enhancement and treated as a separate sub-problem that can be solved by dynamically retrieving a reference image from an external color knowledge base and injecting that reference's global color distribution into a network that predicts only a color residual. Concretely, the module employs a dual-index FAISS retriever over VGG19 global mean and variance vectors, a GlobalSPHistAdaIN injection that reduces a spatial-preserving color histogram to a 512-dimensional vector and modulates features by adaptive instance normalization, and a residual output added to the front end's CIE Lab ab channels. On LOLv1 with CPGA-Net++, the module lowers ΔE2000 from 8.912 to 8.402 and MAEab from 5.265 to 4.712, and color-metric gains appear for LLFormer, FLIGHTNet, and IAT as well. The paper also claims that VGG19-based retrieval beats CLIP-based retrieval for this task across all reported metrics, and that replacing the DIV2K knowledge base with the unseen Flickr2K base retains most of the improvement, meaning the network exploits retrieved statistics rather than memorizing a fixed prior.

Load-bearing premise

The pipeline assumes that a single Top-1 retrieved reference image, chosen by global mean and variance of VGG19 features, carries a color-distribution prior that is useful for the target scene even when reference and target differ in content, and that globally injecting that prior will not damage otherwise correct results; the paper's own results on LOLv2-Synthetic show this assumption can fail.

Editorial extensions

If this is right

  • Color restoration can be evaluated and improved independently of brightness, with dedicated chromatic metrics, and the module can attach to any frozen low-light enhancer.
  • Retrieval for color priors is better served by low- and mid-level texture/structure features than by high-level semantic features such as CLIP.
  • The external knowledge base is interchangeable at inference without retraining: rebuilding the index on Flickr2K retains most of the color-metric gain.
  • Residual prediction is what makes the module generalize across datasets; direct ab prediction loses over 1.7 dB in PSNR on LOLv2 sets.
  • The module's gain scales with the front end's residual color bias: IAT, which leaves the most bias, gains most.

Reading between the lines

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

  • The same retrieve-and-inject pattern could be transferred to other color-governed restoration tasks, such as white balancing, dehazing, or old-photo colorization, where the effective change is only in the knowledge base, the similarity measure, and the injection mechanism.
  • Because the module uses only one Top-1 reference, its quality depends on retrieval accuracy; a controlled study that degrades retrieval rank and measures color-metric change would isolate the retriever's contribution from the injection network's.
  • The reported NIQE decline on unpaired data implies a trade-off between chromatic accuracy and natural image statistics, suggesting that color-focused no-reference metrics are needed to evaluate such modules.
  • The visual analysis suggests the module acts as an automatic white-balance and saturation correction guided by scene-similar reference colors, which could be tested by measuring gray-patch neutrality before and after the module.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 4 minor

Summary. The paper proposes a post-processing color-restoration module for low-light image enhancement. After a frozen front-end LLIE model produces an enhanced image, the module retrieves a reference from a DIV2K FAISS index using global mean and variance of VGG19 features, computes a global SPHist color vector, injects it through GlobalSPHistAdaIN into a DenseNet-121 U-Net, and predicts a residual in Lab. Training uses L1, EMD histogram (against both ground truth and retrieved reference), SSIM, and perceptual losses. The paper reports experiments on LOLv1, LOLv2-Real, LOLv2-Synthetic and unpaired sets with four front ends, plus ablations of retrieval, residual learning, the injection module, and knowledge-base interchangeability.

Significance. The decoupling of color correction from brightness enhancement and the use of an external retrieved color prior is a plausible and potentially useful idea, and the paper contains several exemplary empirical checks: the random-reference and retrieval-free controls, the direct-versus-residual generalization result, the Flickr2K interchangeability test, and the honest disclosure in Section 4.2.3 that the module does not help on LOLv2-Synthetic. If the claims are restricted to real-capture datasets, the contribution is publishable. The main weakness is that the headline "consistently improves" claim is not supported by the paper's own Table 3, and the absence of error bars or significance tests undercuts the small reported gains.

major comments (4)
  1. [Abstract and Section 5] The abstract's claim that the module "consistently improves color-specific metrics" across LOLv1, LOLv2-Real, and LOLv2-Synthetic is contradicted by Table 3: on LOLv2-Synthetic with CPGA-Net++, the module worsens Delta_E2000 from 6.693 to 6.802 and MAEab from 2.813 to 2.998, and also lowers SSIM from 0.920 to 0.913. Section 4.2.3 and the first limitation in Section 5 disclose this, but the conclusion states "Systematic evaluation on three paired datasets bears this out" without qualification. Please restrict the headline claims to real-capture datasets or explicitly characterize synthetic-data results as neutral/slightly negative.
  2. [Section 4.3.4 and Eqs. (24)-(25)] The retrieval-path ablation in Table 8 removes the FAISS retrieval and GlobalSPHistAdaIN, but the retrieval-free condition also loses the reference-histogram loss Lref_hist of Eq. (24) because there is no reference. Since the "Proposed" row uses both the injected reference and a loss pulling the output histogram toward that same reference, the comparison does not isolate the retrieval contribution. Please add a control that retains a reference-histogram term (for example, a random or fixed reference) or that removes the reference from the loss while keeping the conditioning, so that the contribution of retrieval per se is identified.
  3. [Section 4.2 and Tables 1-3] No error bars or significance tests are reported, and several claimed improvements are very small relative to the test set: LOLv1 has only 15 test pairs, and LLFormer + Proposed gains only 0.06 dB in PSNR and 0.001 in SSIM, while CPGA-Net++ on LOLv2-Real changes Delta_E2000 by only 0.041. Without variance estimates or paired significance tests, the statements "consistently improves" and "stable trend" overstate the evidence for these small-delta cases. Please report per-pair distributions, confidence intervals, or paired significance tests at least for the primary front end, or qualify the claims accordingly.
  4. [Section 3.4.4 and Eq. (24)] The same retrieved reference is used both as conditioning input via GlobalSPHistAdaIN and as the target of the auxiliary histogram loss Lref_hist in Eq. (25). This self-referential objective makes the output histogram match the injected distribution by construction, so the effectiveness of the retrieval mechanism may be overestimated. The random-reference ablation in Table 5 partially mitigates this concern, but the paper should explicitly discuss the confounding role of Lref_hist and, ideally, ablate by training without this term or by using a reference that is not injected.
minor comments (4)
  1. [Sections 3.3 and 5] The conclusion describes the structure index as a "variance-covariance index," but Section 3.3 only computes the per-channel variance vector in Eq. (5); no covariance is used. Please align the terminology.
  2. [Table 10] The four-decimal reporting for Delta_E2000 and MAEab implies a precision that a single run cannot support (for example, values differing in the third decimal across retrieval strategies). Round to two decimals or add uncertainty estimates.
  3. [Section 4.2.3] The text says PSNR and SSIM "even dip slightly" under CPGA-Net++, but it should also state that the color metrics worsen in the same row (Delta_E2000 increases by 0.109 and MAEab by 0.185), because the surrounding narrative "the module gains least" could be misread as only brightness metrics being affected.
  4. [Section 4.5] The statement that the GlobalSPHistAdaIN modules together are "1.97 M (2.4%)" should specify whether the percentage is relative to trainable parameters or to the total system parameters, since the latter includes the frozen front end and VGG19 extractor.

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity: the central claim is supported by external benchmarks and ablations, with only a disclosed training-loss coupling to the retrieved reference that is not a reduction by construction.

full rationale

The paper's derivation chain is self-contained against external benchmarks. The color-restoration module is trained with L1, histogram (Eqs. 23-25), SSIM, and perceptual losses against ground truth; the retrieved DIV2K reference is a conditioning input, not the evaluation target. The only self-referential element is Lref_hist (Eqs. 24-25), which nudges the output histogram toward the same reference whose global color vector is injected via GlobalSPHistAdaIN. This is a disclosed training coupling, not a definitional reduction: at inference the reference is selected by FAISS retrieval, and the metrics (∆E2000, MAEab) are computed against ground truth, so improvement is not forced by construction. Evidence against circularity includes the random-reference ablation (Table 5: ∆E2000 improves by 3.4% even with a random reference), the retrieval-free control (Table 8: worse than baseline on all five metrics), the Flickr2K interchangeability test (Table 9: most gains retained with an unseen knowledge base), and the residual-versus-direct comparison (Table 6), which shows the network is not simply copying reference colors. The conclusion that VGG19 retrieval beats CLIP is an empirical comparison on held-out LOLv1 test pairs, not an imported uniqueness claim. Self-citations to CPGA-Net/CPGA-Net++ are not load-bearing because the module is also evaluated with LLFormer, FLIGHTNet, and IAT. The abstract's 'consistently improves' claim is internally contradicted by the paper's own Table 3 on LOLv2-Synthetic with CPGA-Net++ (∆E2000 worsens from 6.693 to 6.802, MAEab from 2.813 to 2.998), and the NIQE results in Table 4 worsen on unpaired data; these are correctness and consistency risks, not circularity. Section 5 acknowledges these limitations explicitly. No circular step meets the evidentiary bar of Eq. X = Eq. Y by construction, or a fitted parameter renamed as a prediction.

Assumptions & free parameters 8 free parameters · 5 assumptions · 0 invented entities

The central claim rests on many hand-chosen hyperparameters and several domain assumptions, the most fragile being that a single DIV2K reference's color distribution is an appropriate prior for any target scene. No new physical entities are introduced; GlobalSPHistAdaIN, the color vector, and the dual-index are architectural or algorithmic components with no independent falsifiable handle outside the paper.

free parameters (8)
  • Retrieval weights alpha and beta = alpha=beta=0.5
    Weighted combination of texture and structure similarity in Eq. 8; chosen by hand and ablated on the LOLv1 test set in Table 10.
  • SPHist Gaussian width sigma = 0.01
    Soft binning width in Eq. 11; chosen by hand and controls the extracted color vector.
  • SPHist bin count K = 256 per ab channel
    Histogram resolution for the color vector and EMD loss; fixed by hand.
  • Histogram loss weight lambda_hist = 2.0
    Loss weight in Eq. 29; chosen by hand with no sensitivity analysis.
  • L1 loss weight lambda_L1 = 0.5
    Per-pixel chromatic L1 loss weight in Eq. 29; chosen by hand.
  • SSIM loss weight lambda_SSIM = 0.1
    Structural loss weight in Eq. 29; chosen by hand.
  • Perceptual loss weight lambda_perc = 0.02
    Perceptual loss weight in Eq. 29; chosen by hand.
  • Top-k references = 1
    Number of retrieved references used for color injection; chosen by hand and acknowledged as a limitation in the conclusion.
assumptions (5)
  • standard math CIE Lab space separates luminance from chroma, and Euclidean distance in Lab approximates perceived color difference.
    Basis for all color losses and metrics; cited via CIE [17] and used throughout the method.
  • domain assumption The retrieved reference image's global color distribution is a useful prior for correcting the target's residual color bias.
    Core RAG premise; tested by ablations in §4.3, but contradicted on LOLv2-Synthetic in Table 3.
  • domain assumption VGG19 intermediate feature global mean and variance capture textural and structural similarity relevant to color.
    Inherited from Pik-Fix [24]; used to build the dual index in §3.3 without independent verification.
  • domain assumption Fixing the front end's L channel and correcting only ab preserves brightness and structure.
    Core modeling choice in §3.4.4 and Eq. 19; if the front end's luminance is wrong, the module cannot fix it.
  • domain assumption Global AdaIN modulation can inject an external color prior without spatial correspondence.
    Design basis of GlobalSPHistAdaIN in §3.4.3; supported only by the paper's own ablations.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Retrieval-Augmented Generation-Based Color Restoration for Low-Light Image Enhancement." pith.science (2026). https://pith.science/paper/I67LXAKH

@misc{pith2026260808211,
  author       = {Pith},
  title        = {Pith review of: Retrieval-Augmented Generation-Based Color Restoration for Low-Light Image Enhancement},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/I67LXAKH}},
  note         = {Machine review of arXiv:2608.08211}
}
read the original abstract

Recent low-light image enhancement (LLIE) methods have driven brightness and structural fidelity close to that of normally-exposed images, yet their outputs still exhibit systematic color shifts such as greenish skies, yellowish faces, and warm-tinted white objects. We attribute this to end-to-end LLIE training coupling brightness, structure, and color within a single network, leaving the color channels weakly supervised. We recast color restoration as an independent sub-problem and decouple it from brightness enhancement, realizing it as a general-purpose post-processing module built on retrieval-augmented generation (RAG). Rather than relying solely on parametric color priors learned during training, the module dynamically retrieves a reference image from an external high-quality color knowledge base and injects its color distribution into a color-restoration network to correct residual bias. The design has three components: (i) a dual-index FAISS retriever built on intermediate VGG19 features, capturing textural and structural similarity through global mean and variance statistics; (ii) GlobalSPHistAdaIN, which reduces the reference spatial-preserving color histogram to a global color vector and modulates network features via adaptive instance normalization, removing dependence on pixel-level correspondence; and (iii) a residual formulation that predicts a color correction over the front-end output. Across LOLv1, LOLv2-Real, and LOLv2-Synthetic, the module consistently improves color-specific metrics, and it remains effective when the front end is swapped among CPGA-Net++, LLFormer, FLIGHTNet, and IAT, confirming cross-front-end generality. Ablations show that a VGG19 dual index outperforms CLIP-based retrieval, indicating that color restoration depends on textural and structural similarity rather than high-level semantics.

Figures

Figures reproduced from arXiv: 2608.08211 by the authors.

Figure 1
Figure 1. Overall system architecture. Stage 1 (frozen LLIE front end) produces a brightness-enhanced image; Stage 2 retrieves a reference from an external knowledge base and injects its color distribution to predict a color residual. 8 [PITH_FULL_IMAGE:figures/full_fig_p008_1.png] view at source ↗
Figure 2
Figure 2. Dual-index construction: VGG19 features of each knowledge-base image are reduced to global mean and variance vectors, stored in a texture index and a structure index respectively. 10 [PITH_FULL_IMAGE:figures/full_fig_p010_2.png] view at source ↗
Figure 3
Figure 3. Reference retrieval and weighted re-ranking. Each index returns candidates; scores are merged and re-ranked by α, β to select the Top-1 reference. 11 [PITH_FULL_IMAGE:figures/full_fig_p011_3.png] view at source ↗
Figures from the paper (8 more)
Figure 4
Figure 4. Figure 4: Color-restoration module: SPHist color-vector extraction from the retrieved reference, GlobalSPHistAdaIN injection into encoder features, and residual color prediction. 12 [PITH_FULL_IMAGE:figures/full_fig_p012_4.png]
Figure 5
Figure 5. Figure 5: Overall architecture of the color-restoration (colorization) network. The Dense-121 U-Net encoder injects the reference color prior through a GlobalSPHistAdaIN module after each transition layer; the decoder progressively restores resolution via skip connections and ou…
Figure 6
Figure 6. Figure 6: Spatial-preserving color histogram (SPHist) computation via differentiable Gaussian soft assignment over the a, b chromatic channels. 15 [PITH_FULL_IMAGE:figures/full_fig_p015_6.png]
Figure 7
Figure 7. Figure 7: GlobalSPHistAdaIN modulation flow: the 512-dimensional color vector generates per-channel scale and shift parameters applied to instance-normalized features. Its core mechanism is AdaIN [13]: given a feature map f with scale γ and shift β, AdaIN(f, γ, β) = γ ·  f − µ(…
Figure 8
Figure 8. Figure 8: Visual comparison on LOLv1. Top: low-light input and the four front-end outputs; bottom: ground truth and each front end paired with the proposed module. LOLv2 comparison. Figures 9 and 10 show LOLv2-Real [33] and LOLv2-Synthetic [33]. On real captures the module corre…
Figure 9
Figure 9. Figure 9: Visual comparison on LOLv2-Real Captured across CPGA-Net++ and FLIGHTNet front ends with and without the proposed module [PITH_FULL_IMAGE:figures/full_fig_p024_9.png]
Figure 10
Figure 10. Figure 10: Visual comparison on LOLv2-Synthetic; improvements are smaller due to the mismatch between synthetic color statistics and the natural-image knowledge base. 4.5 Efficiency and Complexity The system totals about 85.74 M parameters, of which the frozen CPGA-Net++ [6] (0.…
Figure 11
Figure 11. Figure 11: Retrieval–restoration correspondence. Left: bookshelf scene, where a warm-dominant reference guides recovery of warm-brown tones. Right: near-neutral storage scene, where a low-saturation reference yields a near-zero residual. (a) ground truth; (b) front-end output; (…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

40 extracted references · 34 canonical work pages

  1. [1]

    The retinex theory of color vision,

    E. H. Land, “The retinex theory of color vision,”Scientific American, vol. 237, no. 6, pp. 108–129, 1977

    work page 1977

  2. [2]

    LIME: Low-light image enhancement via illumination map estimation,

    X. Guo, Y . Li, and H. Ling, “LIME: Low-light image enhancement via illumination map estimation,” IEEE Trans. Image Process., vol. 26, no. 2, pp. 982–993, 2016

    work page 2016

  3. [3]

    Deep retinex decomposition for low-light enhancement,

    C. Wei, W. Wang, W. Yang, and J. Liu, “Deep retinex decomposition for low-light enhancement,”arXiv preprint arXiv:1808.04560, 2018

    arXiv 2018

  4. [4]

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

    T. Wang, K. Zhang, T. Shen, W. Luo, B. Stenger, and T. Lu, “Ultra-high-definition low-light image enhancement: A benchmark and transformer-based method,” inProc. AAAI Conf. Artificial Intelligence, 2023, vol. 37, no. 3, pp. 2654–2662

    work page 2023

  5. [5]

    A lightweight low-light image enhancement network via channel prior and gamma correction,

    S.-E. Weng, S.-G. Miaou, and R. Christanto, “A lightweight low-light image enhancement network via channel prior and gamma correction,”arXiv preprint arXiv:2402.18147, 2024

    work page arXiv 2024

  6. [6]

    Rethinking Theoretical Illumination for Efficient Low-Light Image Enhancement

    S.-E. Weng et al., “Rethinking theoretical illumination for efficient low-light image enhancement,”arXiv preprint arXiv:2409.05274, 2024

    work page Pith review arXiv 2024

  7. [7]

    Flight mode on: A feather-light network for low-light image enhancement,

    M. Ozcan, H. Ergezer, and M. Ayazo ˘glu, “Flight mode on: A feather-light network for low-light image enhancement,” inProc. IEEE/CVF Conf. Computer Vision and Pattern Recognition, 2023, pp. 4226–4235

    work page 2023

  8. [8]

    You only need 90k parameters to adapt light: a lightweight transformer for image enhancement and exposure correction,

    Z. Cui et al., “You only need 90k parameters to adapt light: a lightweight transformer for image enhancement and exposure correction,”arXiv preprint arXiv:2205.14871, 2022

    arXiv 2022

Show all 40 references
  1. [9]

    Retrieval-augmented generation for knowledge-intensive NLP tasks,

    P. Lewis et al., “Retrieval-augmented generation for knowledge-intensive NLP tasks,”Advances in Neural Information Processing Systems, vol. 33, pp. 9459–9474, 2020

    2020

  2. [10]

    The unreasonable effectiveness of deep features as a perceptual metric,

    R. Zhang, P. Isola, A. A. Efros, E. Shechtman, and O. Wang, “The unreasonable effectiveness of deep features as a perceptual metric,” inProc. IEEE Conf. Computer Vision and Pattern Recognition, 2018, pp. 586–595

    2018

  3. [11]

    Billion-scale similarity search with GPUs,

    J. Johnson, M. Douze, and H. Jégou, “Billion-scale similarity search with GPUs,”IEEE Trans. Big Data, vol. 7, no. 3, pp. 535–547, 2019

    2019

  4. [12]

    Very deep convolutional networks for large-scale image recognition,

    K. Simonyan and A. Zisserman, “Very deep convolutional networks for large-scale image recognition,” arXiv preprint arXiv:1409.1556, 2014

    2014 arXiv

  5. [13]

    Arbitrary style transfer in real-time with adaptive instance normalization,

    X. Huang and S. Belongie, “Arbitrary style transfer in real-time with adaptive instance normalization,” inProc. IEEE Int. Conf. Computer Vision, 2017, pp. 1501–1510

    2017

  6. [14]

    Learning transferable visual models from natural language supervision,

    A. Radford et al., “Learning transferable visual models from natural language supervision,” inProc. Int. Conf. Machine Learning, 2021, pp. 8748–8763

    2021

  7. [15]

    Contrast limited adaptive histogram equalization,

    K. Zuiderveld, “Contrast limited adaptive histogram equalization,” inGraphics Gems IV, 1994, pp. 474– 485

    1994

  8. [16]

    A multiscale retinex for bridging the gap between color images and the human observation of scenes,

    D. J. Jobson, Z. Rahman, and G. A. Woodell, “A multiscale retinex for bridging the gap between color images and the human observation of scenes,”IEEE Trans. Image Process., vol. 6, no. 7, pp. 965–976, 1997. 27

    1997

  9. [17]

    Colorimetry — Part 4: CIE 1976 L*a*b* Colour Space, CIE 015:2018, International Commission on Illumination, Vienna, Austria, 2019

    1976

  10. [18]

    Let there be color! Joint end-to-end learning of global and local image priors for automatic image colorization with simultaneous classification,

    S. Iizuka, E. Simo-Serra, and H. Ishikawa, “Let there be color! Joint end-to-end learning of global and local image priors for automatic image colorization with simultaneous classification,”ACM Trans. Graphics, vol. 35, no. 4, pp. 1–11, 2016

    2016

  11. [19]

    Colorful image colorization,

    R. Zhang, P. Isola, and A. A. Efros, “Colorful image colorization,” inProc. European Conf. Computer Vision, 2016, pp. 649–666

    2016

  12. [20]

    Transferring color to greyscale images,

    T. Welsh, M. Ashikhmin, and K. Mueller, “Transferring color to greyscale images,” inProc. 29th Annu. Conf. Computer Graphics and Interactive Techniques, 2002, pp. 277–280

    2002

  13. [21]

    Colorization by example,

    R. Ironi, D. Cohen-Or, and D. Lischinski, “Colorization by example,”Rendering Techniques, vol. 29, pp. 201–210, 2005

    2005

  14. [22]

    Image colorization using similar images,

    R. K. Gupta, A. Y .-S. Chia, D. Rajan, E. S. Ng, and H. Zhiyong, “Image colorization using similar images,” inProc. 20th ACM Int. Conf. Multimedia, 2012, pp. 369–378

    2012

  15. [23]

    Deep exemplar-based colorization,

    M. He, D. Chen, J. Liao, P. V . Sander, and L. Yuan, “Deep exemplar-based colorization,”ACM Trans. Graphics, vol. 37, no. 4, pp. 1–16, 2018

    2018

  16. [24]

    Pik-Fix: Restoring and colorizing old photos,

    R. Xu et al., “Pik-Fix: Restoring and colorizing old photos,” inProc. IEEE/CVF Winter Conf. Applica- tions of Computer Vision, 2023, pp. 1724–1734

    2023

  17. [25]

    Deep residual learning for image recognition,

    K. He, X. Zhang, S. Ren, and J. Sun, “Deep residual learning for image recognition,” inProc. IEEE Conf. Computer Vision and Pattern Recognition, 2016, pp. 770–778

    2016

  18. [26]

    U-Net: Convolutional networks for biomedical image segmentation,

    O. Ronneberger, P. Fischer, and T. Brox, “U-Net: Convolutional networks for biomedical image segmentation,” inProc. Int. Conf. Medical Image Computing and Computer-Assisted Intervention, 2015, pp. 234–241

    2015

  19. [27]

    Color indexing,

    M. J. Swain and D. H. Ballard, “Color indexing,”Int. J. Computer Vision, vol. 7, no. 1, pp. 11–32, 1991

    1991

  20. [28]

    Instance normalization: The missing ingredient for fast stylization,

    D. Ulyanov, A. Vedaldi, and V . Lempitsky, “Instance normalization: The missing ingredient for fast stylization,”arXiv preprint arXiv:1607.08022, 2016

    2016 arXiv

  21. [29]

    A survey of retrieval-augmented generation (RAG) for large language models,

    Y . Ma et al., “A survey of retrieval-augmented generation (RAG) for large language models,” inProc. 2025 Int. Conf. Trustworthy Big Data and Artificial Intelligence (ICTBAI), 2025, pp. 7–13

    2025

  22. [30]

    Semi-parametric neural image synthesis,

    A. Blattmann, R. Rombach, K. Oktay, J. Müller, and B. Ommer, “Semi-parametric neural image synthesis,”arXiv preprint arXiv:2204.11824, 2022

    2022 arXiv

  23. [31]

    NTIRE 2017 challenge on single image super-resolution: Dataset and study,

    E. Agustsson and R. Timofte, “NTIRE 2017 challenge on single image super-resolution: Dataset and study,” inProc. IEEE Conf. Computer Vision and Pattern Recognition Workshops, 2017, pp. 126–135

    2017

  24. [32]

    Densely connected convolutional networks,

    G. Huang, Z. Liu, L. Van Der Maaten, and K. Q. Weinberger, “Densely connected convolutional networks,” inProc. IEEE Conf. Computer Vision and Pattern Recognition, 2017, pp. 4700–4708

    2017

  25. [33]

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

    W. Yang, W. Wang, H. Huang, S. Wang, and J. Liu, “Sparse gradient regularized deep retinex network for robust low-light image enhancement,”IEEE Trans. Image Process., vol. 30, pp. 2072–2086, 2021

    2021

  26. [34]

    Naturalness preserved enhancement algorithm for non-uniform illumination images,

    S. Wang, J. Zheng, H.-M. Hu, and B. Li, “Naturalness preserved enhancement algorithm for non-uniform illumination images,”IEEE Trans. Image Process., vol. 22, no. 9, pp. 3538–3548, 2013. 28

    2013

  27. [35]

    Perceptual quality assessment for multi-exposure image fusion,

    K. Ma, K. Zeng, and Z. Wang, “Perceptual quality assessment for multi-exposure image fusion,”IEEE Trans. Image Process., vol. 24, no. 11, pp. 3345–3356, 2015

    2015

  28. [36]

    On the evaluation of illumination compensation algorithms,

    V . V onikakis, R. Kouskouridas, and A. Gasteratos, “On the evaluation of illumination compensation algorithms,”Multimedia Tools and Applications, vol. 77, no. 8, pp. 9211–9231, 2018

    2018

  29. [37]

    Contrast enhancement based on layered difference representation of 2D histograms,

    C. Lee, C. Lee, and C.-S. Kim, “Contrast enhancement based on layered difference representation of 2D histograms,”IEEE Trans. Image Process., vol. 22, no. 12, pp. 5372–5384, 2013

    2013

  30. [38]

    The development of the CIE 2000 colour-difference formula: CIEDE2000,

    M. R. Luo, G. Cui, and B. Rigg, “The development of the CIE 2000 colour-difference formula: CIEDE2000,”Color Research & Application, vol. 26, no. 5, pp. 340–350, 2001

    2000

  31. [39]

    Making a ‘completely blind’ image quality analyzer,

    A. Mittal, R. Soundararajan, and A. C. Bovik, “Making a ‘completely blind’ image quality analyzer,” IEEE Signal Process. Lett., vol. 20, no. 3, pp. 209–212, 2012

    2012

  32. [40]

    Enhanced deep residual networks for single image super-resolution,

    B. Lim, S. Son, H. Kim, S. Nah, and K. M. Lee, “Enhanced deep residual networks for single image super-resolution,” inProc. IEEE Conf. Computer Vision and Pattern Recognition Workshops, 2017, pp. 136–144. 29

    2017

Pith tools

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