Pith. sign in

REVIEW 3 major objections 6 minor 34 references

UDBE: Unsupervised Diffusion-based Brightness Enhancement in Underwater Images

T0 review · 3 major / 6 minor · reviewed 2026-08-10 · deepseek-v4-flash

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

Pith's one-line read UDBE uses an unsupervised conditional diffusion process to brighten underwater images without paired references.

desk verdict The method is a reasonable adaptation of CLE Diffusion to underwater brightness, but the evaluation only measures inversion of the authors' own synthetic brightness shift, and Table I contradicts the text on RUIE, so the central claim is unsupported. read the letter →

arxiv 2501.16211 v1 pith:WI4XDEJF submitted 2025-01-27 cs.CV cs.AIeess.IV

Tatiana Taís Schein , Gustavo Pereira de Almeira , Stephanie Loi Brião , Rodrigo Andrade de Bem , Felipe Gomes de Oliveira , Paulo L. J. Drews-Jr This is my paper · ORCID
classification cs.CVcs.AIeess.IV
keywords underwaterimageenhancementbrightnessdiffusionmodelunsupervisedlearningconditionalDDPMqualityassessmentSNRmap
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

The paper proposes UDBE, an unsupervised diffusion method for increasing brightness in underwater images without paired reference images. It creates synthetic training pairs by subtracting or adding random brightness values in the 50–100 range to raw underwater images, then trains a DDPM-based network to reverse the darkening. Color maps and signal-to-noise maps are appended to the input so the model brightens without distorting color. On the UIEB, SUIM, and RUIE benchmarks, the paper reports higher PSNR, SSIM, UIQM, and UISM scores than three prior enhancement methods, alongside visually clearer output. If correct, this would make diffusion-based brightness enhancement practical for underwater datasets that lack ground-truth bright versions.

What carries the argument

The central mechanism is the conditional reverse diffusion of a DDPM, steered by three conditioning signals. A FiLM layer (feature-wise linear modulation) in every U-Net layer applies an affine transformation driven by the desired brightness level; the network input also concatenates a color map, in which each RGB channel is normalized by its maximum value, and an SNR map, defined as the Gaussian-blurred image divided by the absolute difference between the image and its blur, which highlights dark and noisy regions. These signals let one network brighten an image while keeping colors stable. The training objective is a weighted sum of LPIPS, SSIM, MSE, brightness, and color losses, with the brightness and color terms added after the first twenty epochs.

What would settle it

Collect a set of real underwater scenes photographed in darkness and then re-photographed under controlled artificial lighting to form true bright references; run UDBE on the dark versions and compute PSNR/SSIM against those true references. If the scores are no better than a simple histogram stretch or than inverting the uniform shift, the synthetic training assumption is the limiting factor; if they are close to the paper's reported values, the assumption transfers.

Watch

Extended reading notes

Core claim

The central claim is that brightness degradation in underwater images can be learned and reversed in a fully unsupervised way by a conditional diffusion model. For each raw image, the method constructs a low-brightness input by subtracting a random value between 50 and 100 and a high-brightness reference by adding a random value between 50 and 100. The network, a U-Net inside a DDPM with FiLM brightness conditioning, is trained with a weighted mix of LPIPS, SSIM, MSE, grayscale brightness, and angular color losses. During inference, the raw dark image is concatenated with a normalized color map and an SNR map, and the reverse diffusion process produces the brightened result. The paper reports that UDBE surpasses RUIDL, UDNet, and UESAM in PSNR and SSIM on all three datasets, and in most UIQM and UISM comparisons, and it presents this as the first diffusion-based brightness enhancement for underwater images.

Load-bearing premise

The method assumes that dark underwater images can be simulated by subtracting a uniform brightness value between 50 and 100 from an already captured raw image, and that adding 50 to 100 produces a correct bright reference; both training and evaluation depend on that synthetic brightness shift matching real underwater darkness.

Editorial extensions

If this is right

  • If the claim is correct, underwater enhancement no longer requires paired reference images, so any raw underwater collection can serve as training data.
  • The same conditioning recipe—brightness level, color map, and SNR map—could be reused for other unpaired restoration tasks where color and noise structure are known.
  • Diffusion-based restoration would become competitive with established CNN and self-supervised baselines on standard underwater benchmarks, at least for brightness.
  • Because inference uses DDIM sampling, the enhanced output is produced without the full iterative cost of a standard diffusion sampler.

Reading between the lines

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

  • The reported full-reference scores measure how well the model inverts a uniform additive brightening of the same image, so they should not be read as direct evidence about real depth-induced attenuation; testing on true paired underwater captures is the natural next step.
  • The paper's own table shows that on RUIE the non-reference metrics UIQM and UISM are lower for UDBE than for RUIDL, so the 'outperforms' claim is strongest for full-reference similarity on synthetic brightness shifts and is not uniform across every metric.
  • A testable extension is to replace the uniform brightness augmentation with a wavelength-dependent attenuation and scattering model; the change in benchmark scores would separate what UDBE learns about brightness inversion from what it learns about real underwater physics.
  • The color-map/SNR-map conditioning is a general mechanism that could transfer to other unpaired image restoration problems, such as low-light photography or turbid-water video, where color and noise are the main degradation channels.
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

3 major / 6 minor

Summary. The paper proposes UDBE, an unsupervised diffusion-based method for brightness enhancement of underwater images. The method constructs synthetic high- and low-brightness versions of raw underwater images by adding or subtracting random brightness values, conditions a DDPM on color maps and SNR maps, and compares against RUIDL, UDNet, and UESAM on UIEB, SUIM, and RUIE using PSNR, SSIM, UIQM, and UISM. The authors claim that UDBE outperforms the comparison methods and is robust across the three datasets.

Significance. If the central claim were supported, the paper would offer a practically useful unsupervised diffusion-based alternative for underwater brightness enhancement, and the released code would aid reproducibility. The idea of conditioning a diffusion model on color and SNR maps for brightness control is appealing and the qualitative figures show some plausible improvements. However, the current evaluation does not establish the claimed superiority: the full-reference metrics are computed against synthetic references generated by the authors' own brightness operation, and the non-reference metrics on RUIE contradict the stated conclusion. The significance of the contribution therefore cannot be assessed from the evidence as printed.

major comments (3)
  1. [Section III-B-1 and Section IV-E] The quantitative evaluation is self-referential. Both the training data and the test references are generated from the same raw images by adding or subtracting uniform random brightness values in [50,100] and [-50,-100], respectively. Section IV-E states that the generated high-brightness images are used as references for PSNR and SSIM and the low-brightness images as inputs. A model trained to invert this exact synthetic brightness shift can therefore score high on PSNR and SSIM without performing the physical operations that characterize real underwater darkness, such as wavelength-dependent attenuation, scattering, and artificial-lighting artifacts. Since UIEB provides real reference images, the authors had the option to evaluate on those; the choice of synthetic references makes the reported full-reference numbers non-representative of real enhancement.
  2. [Table I, RUIE rows, and Section IV-E] The text asserts that 'our approach outperforms the comparison methods' and the abstract claims robustness on UIQM and UISM, but Table I shows UDBE's RUIE UIQM (0.4531) is below RUIDL (0.829) and UDNet (0.624), and its UISM (3.902) is below RUIDL (4.674) and UDNet (4.005). This is an internal contradiction: either the table or the conclusion is wrong, so the superiority claim cannot be accepted as printed.
  3. [Section III-C and Eqs. (11)-(15)] There is a mismatch between the stated reference data and the loss functions. The text says that high-brightness images H are used as reference data during the unsupervised learning process, but all loss equations compare the enhanced output B to the raw image I (Eqs. 11-15). If H is the intended target, the losses are incorrectly specified; if I is the target, the method is not optimizing toward the stated brightness-enhancement goal. This affects what the model actually learns. In addition, the method is described as unsupervised, but Section III-B/III-C constructs explicit paired synthetic data (L, H) from I and uses H as reference; the terminology should be justified or revised.
minor comments (6)
  1. [Section III-B-1] The low-brightness range is stated as [-50, -100]; since these values are said to be subtracted from the raw images, the sign convention makes the operation add brightness. The intended range is likely [50, 100].
  2. [Abstract and Section V] The abstract promises an 'impressive accuracy rate' and 'robustness' on UIQM and UISM, while the conclusion says 'reasonable performance'; these characterizations should be aligned with the quantitative results.
  3. [Section III-C] The acronym 'CLE' in 'CLE Diffusion' is not defined, and the relationship to reference [24] should be stated explicitly.
  4. [Section IV-A] The dataset descriptions omit that SUIM is a semantic segmentation benchmark and that RUIE contains multiple evaluation subsets; the exact splits and subsets used should be reported.
  5. [Section IV-E and Figure 2] The text discusses 'PNSR' instead of 'PSNR', and the qualitative analysis would be more convincing if accompanied by quantitative brightness or illumination statistics in addition to subjective visual inspection.
  6. [Eq. (13)] The angular color loss notation is unclear; the angle operation between the three-dimensional RGB vectors should be defined explicitly.

Circularity Check

1 steps flagged · score 7.0 of 10

Full-reference PSNR/SSIM evaluation is self-referential: the 'high-brightness reference' is generated by the same synthetic brightness transformation used to build the input, so the headline accuracy measures inversion of the authors' own brightness shift; the remaining UIQM/UISM evidence contradicts the claimed superiority.

  1. self definitional [Section III-B-1 (Brightness Adjustment) and Section IV-E (Quantitative Evaluation), Table I]
    "For full reference analysis, through PNSR and SSIM metrics, the generated high-brightness underwater images are considered as reference images, and the low-brightness underwater images as input images."

    The reference used for PSNR/SSIM is not independent: Section III-B-1 constructs H as raw images plus random values in [50,100] and L as raw images minus random values in [50,100], and Section IV-E evaluates L as input and H as reference. Thus the full-reference metrics measure inversion of the authors' own synthetic additive brightness shift, not enhancement of real underwater attenuation, scattering, or artificial-lighting artifacts. Since the model is trained on the same H/L construction (Section III-C), the 'prediction' is the inverse of the transformation that defines the reference; PSNR/SSIM are self-referential and cannot support the abstract's 'impressive accuracy rate' or Section IV-E's claim to outperform comparison methods.

full rationale

The central empirical claim (abstract: 'impressive accuracy rate'; Section IV-E: 'our approach outperforms the comparison methods') rests on Table I. The PSNR and SSIM columns are self-referential: Section III-B-1 defines the high-brightness reference H as raw I plus uniform random values in [50,100] and the low-brightness input L as raw I minus uniform random values in [50,100], and Section IV-E then scores the model on L->H with H as reference. The model is trained on the same synthetic transformation (Section III-C), so the full-reference metrics report success at inverting the authors' own brightness shift; they say nothing about real underwater degradation (wavelength-dependent attenuation, scattering, artificial light). The non-reference UIQM/UISM metrics are not circular, but they do not rescue the claim: on RUIE, Table I lists UDBE UIQM=0.4531 and UISM=3.902 versus RUIDL UIQM=0.829 and UISM=4.674, directly contradicting Section IV-E's statement that UDBE outperforms the comparison methods. I also note that two of the three baselines (RUIDL [5], UESAM [19]) are co-authored by members of this paper's team, but that fact is not itself the circularity; the circularity is the synthetic input/reference construction that makes the headline PSNR/SSIM numbers a measure of transformation inversion rather than underwater enhancement. Because the primary quantitative evidence reduces by construction to the inverse of a self-defined synthetic shift, the score is elevated despite the non-circular (and internally inconsistent) UIQM/UISM results.

Assumptions & free parameters 2 free parameters · 2 assumptions · 0 invented entities

The central claim leans on two hand-picked parameter sets (brightness shift ranges, loss weights) and on the unvalidated assumption that uniform brightness addition and subtraction captures underwater lighting degradation. No new physical or architectural entities are postulated.

free parameters (2)
  • Brightness shift ranges for synthetic H and L = H: +[50,100]; L: -[50,100]
    Hand-chosen constants define both the training target and the evaluation reference. No sensitivity analysis is given.
  • Loss weights gamma_1..gamma_5 = 30, 2.83, 1, 20, 100
    Explicitly described as 'empirically defined' in Section III-C-2 to improve brightness enhancement, meaning they are tuned on the benchmark task.
assumptions (2)
  • ad hoc to paper Adding or subtracting a uniform random brightness value to a raw underwater image produces realistic low- and high-brightness versions of that image.
    Section III-B-1 defines H and L this way; the entire training signal and evaluation reference depend on this equivalence. The paper provides no evidence that this synthetic transformation matches real underwater lighting degradation.
  • domain assumption ImageNet-pretrained AlexNet features used in LPIPS remain meaningful for underwater images.
    Section III-C-2 uses LPIPS with AlexNet trained on ImageNet; no underwater-specific validation is given.

how reviews work

0 comments
Cite this review

Pith. "Pith review of UDBE: Unsupervised Diffusion-based Brightness Enhancement in Underwater Images." pith.science (2026). https://pith.science/paper/WI4XDEJF

@misc{pith2026250116211,
  author       = {Pith},
  title        = {Pith review of: UDBE: Unsupervised Diffusion-based Brightness Enhancement in Underwater Images},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/WI4XDEJF}},
  note         = {Machine review of arXiv:2501.16211}
}
read the original abstract

Activities in underwater environments are paramount in several scenarios, which drives the continuous development of underwater image enhancement techniques. A major challenge in this domain is the depth at which images are captured, with increasing depth resulting in a darker environment. Most existing methods for underwater image enhancement focus on noise removal and color adjustment, with few works dedicated to brightness enhancement. This work introduces a novel unsupervised learning approach to underwater image enhancement using a diffusion model. Our method, called UDBE, is based on conditional diffusion to maintain the brightness details of the unpaired input images. The input image is combined with a color map and a Signal-Noise Relation map (SNR) to ensure stable training and prevent color distortion in the output images. The results demonstrate that our approach achieves an impressive accuracy rate in the datasets UIEB, SUIM and RUIE, well-established underwater image benchmarks. Additionally, the experiments validate the robustness of our approach, regarding the image quality metrics PSNR, SSIM, UIQM, and UISM, indicating the good performance of the brightness enhancement process. The source code is available here: https://github.com/gusanagy/UDBE.

Figures

Figures reproduced from arXiv: 2501.16211 by the authors.

Figure 1
Figure 1. Overview of the proposed methodology for brightness restoration in underwater images. The proposed approach is [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Qualitative comparison of restored underwater images on the UIEB (1 [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

34 extracted references · 26 canonical work pages

  1. [1]

    Underwater image restoration: A state-of-the-art review,

    S. Fayaz, S. Parah, G. Qureshi, and V . Kumar, “Underwater image restoration: A state-of-the-art review,” IET Image Processing , vol. 15, 2020

    work page 2020

  2. [2]

    Underwater image enhancement by fusion,

    S. Abbas, “Underwater image enhancement by fusion,” Ph.D. disserta- tion, CAPITAL UNIVERSITY , 2019

    work page 2019

  3. [3]

    Denoising diffusion probabilistic models,

    J. Ho, A. Jain, and P. Abbeel, “Denoising diffusion probabilistic models,” Advances in neural information processing systems , vol. 33, pp. 6840– 6851, 2020

    2020

  4. [4]

    U-net: Convolutional networks for biomedical image segmentation,

    O. Ronneberger, P. Fischer, and T. Brox, “U-net: Convolutional networks for biomedical image segmentation,” in Medical image computing and computer-assisted intervention–MICCAI 2015: 18th international con- ference, Munich, Germany, October 5-9, 2015, proceedings, part III 18 . Springer, 2015, pp. 234–241

    2015

  5. [5]

    Diving into clarity: Restoring underwater images using deep learning,

    L. A. Martinho, J. M. B. Calvalcanti, J. L. Pio, and F. G. Oliveira, “Diving into clarity: Restoring underwater images using deep learning,” Journal of Intelligent & amp; Robotic Systems , vol. 110, no. 1, Feb. 2024. [Online]. Available: http://dx.doi.org/10.1007/ s10846-024-02065-8

    work page 2024

  6. [6]

    Single image haze removal using dark channel prior,

    K. He, J. Sun, and X. Tang, “Single image haze removal using dark channel prior,” IEEE transactions on pattern analysis and machine intelligence, vol. 33, no. 12, pp. 2341–2353, 2010

    2010

  7. [7]

    Underwater image restoration based on image blurriness and light absorption,

    Y .-T. Peng and P. C. Cosman, “Underwater image restoration based on image blurriness and light absorption,” IEEE transactions on image processing, vol. 26, no. 4, pp. 1579–1594, 2017

    work page 2017

  8. [8]

    Generalization of the dark channel prior for single image restoration,

    Y .-T. Peng, K. Cao, and P. C. Cosman, “Generalization of the dark channel prior for single image restoration,” IEEE Transactions on Image Processing, vol. 27, no. 6, pp. 2856–2868, 2018

    work page 2018

Show all 34 references
  1. [9]

    Shallow- water image enhancement using relative global histogram stretching based on adaptive parameter acquisition,

    D. Huang, Y . Wang, W. Song, J. Sequeira, and S. Mavromatis, “Shallow- water image enhancement using relative global histogram stretching based on adaptive parameter acquisition,” in MultiMedia Modeling: 24th International Conference, MMM 2018, Bangkok, Thailand, February 5- 7...

    2018

  2. [10]

    The retinex based improved underwater image enhancement,

    N. Hassan, S. Ullah, N. Bhatti, H. Mahmood, and M. Zia, “The retinex based improved underwater image enhancement,” Multimedia Tools and Applications, vol. 80, pp. 1839–1857, 2021

    2021

  3. [11]

    Bayesian retinex underwater image enhancement,

    P. Zhuang, C. Li, and J. Wu, “Bayesian retinex underwater image enhancement,” Engineering Applications of Artificial Intelligence , vol. 101, p. 104171, 2021

    2021

  4. [12]

    Enhancement of underwater images with retinex transmission map and adaptive color correction,

    E. Chen, T. Ye, Q. Chen, B. Huang, and Y . Hu, “Enhancement of underwater images with retinex transmission map and adaptive color correction,” Applied Sciences, vol. 13, no. 3, p. 1973, 2023

    1973

  5. [13]

    Underwater image enhancement based on fusion of intensity transformation techniques,

    L. A. Martinho, F. G. Oliveira, J. M. B. Cavalcanti, and J. L. S. Pio, “Underwater image enhancement based on fusion of intensity transformation techniques,” in 2022 Latin American Robotics Sympo- sium (LARS), 2022 Brazilian Symposium on Robotics (SBR), and 2022 Workshop on Ro...

    2022

  6. [14]

    Uiecˆ 2-net: Cnn-based under- water image enhancement using two color space,

    Y . Wang, J. Guo, H. Gao, and H. Yue, “Uiecˆ 2-net: Cnn-based under- water image enhancement using two color space,” Signal Processing: Image Communication, vol. 96, p. 116250, 2021

    2021

  7. [15]

    Underwater image enhancement via medium transmission-guided multi-color space embedding,

    C. Li, S. Anwar, J. Hou, R. Cong, C. Guo, and W. Ren, “Underwater image enhancement via medium transmission-guided multi-color space embedding,” IEEE Transactions on Image Processing, vol. 30, pp. 4985– 5000, 2021

    2021

  8. [16]

    Ca- gan: Class-condition attention gan for underwater image enhancement,

    J. Wang, P. Li, J. Deng, Y . Du, J. Zhuang, P. Liang, and P. Liu, “Ca- gan: Class-condition attention gan for underwater image enhancement,” IEEE access, vol. 8, pp. 130 719–130 728, 2020

    2020

  9. [17]

    Underwater image restoration and enhancement via residual two-fold attention networks

    B. Fu, L. Wang, R. Wang, S. Fu, F. Liu, and X. Liu, “Underwater image restoration and enhancement via residual two-fold attention networks.” Int. J. Comput. Intell. Syst. , vol. 14, no. 1, pp. 88–95, 2021

    2021

  10. [18]

    Adaptive uncertainty distribution in deep learning for unsupervised underwater image enhancement,

    A. Saleh, M. Sheaves, D. Jerry, and M. Rahimi Azghadi, “Adaptive uncertainty distribution in deep learning for unsupervised underwater image enhancement,” SSRN Electronic Journal , 2023. [Online]. Available: https://ssrn.com/abstract=4362438

    2023

  11. [19]

    Underwater enhancement based on a self-learning strategy and attention mechanism for high-intensity regions,

    C. D. Mello, B. U. Moreira, P. J. D. de Oliveira Evald, P. J. L. Drews, and S. S. da Costa Botelho, “Underwater enhancement based on a self-learning strategy and attention mechanism for high-intensity regions,” Computers & Graphics, vol. 107, p. 264–276, Oct. 2022. [Online...

    2022 doi

  12. [20]

    Wavelet-based fourier infor- mation interaction with frequency diffusion adjustment for underwater image restoration,

    C. Zhao, W. Cai, C. Dong, and C. Hu, “Wavelet-based fourier infor- mation interaction with frequency diffusion adjustment for underwater image restoration,” arXiv preprint arXiv:2311.16845 , 2023

    2023 arXiv

  13. [21]

    Uiedp: Underwater image enhancement with diffusion prior,

    D. Du, E. Li, L. Si, F. Xu, J. Niu, and F. Sun, “Uiedp: Underwater image enhancement with diffusion prior,” arXiv preprint arXiv:2312.06240 , 2023

    2023 arXiv

  14. [22]

    Underwater image enhancement method based on denoising diffusion probabilistic model,

    S. Lu, F. Guan, H. Zhang, and H. Lai, “Underwater image enhancement method based on denoising diffusion probabilistic model,” Journal of Visual Communication and Image Representation , vol. 96, p. 103926, 2023

    2023

  15. [23]

    Underwater image enhance- ment by transformer-based diffusion model with non-uniform sampling for skip strategy,

    Y . Tang, H. Kawasaki, and T. Iwaguchi, “Underwater image enhance- ment by transformer-based diffusion model with non-uniform sampling for skip strategy,” in Proceedings of the 31st ACM International Con- ference on Multimedia , 2023, pp. 5419–5427

    2023

  16. [24]

    Cle diffusion: Controllable light enhancement diffusion model,

    Y . Yin, D. Xu, C. Tan, P. Liu, Y . Zhao, and Y . Wei, “Cle diffusion: Controllable light enhancement diffusion model,” in Proceedings of the 31st ACM International Conference on Multimedia , 2023, pp. 8145– 8156

    2023

  17. [25]

    Film: Visual reasoning with a general conditioning layer,

    E. Perez, F. Strub, H. De Vries, V . Dumoulin, and A. Courville, “Film: Visual reasoning with a general conditioning layer,” in Proceedings of the AAAI conference on artificial intelligence , vol. 32, no. 1, 2018

    2018

  18. [26]

    Denoising diffusion implicit models,

    J. Song, C. Meng, and S. Ermon, “Denoising diffusion implicit models,” arXiv preprint arXiv:2010.02502 , 2020

    2010 arXiv

  19. [27]

    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,” in Proceedings of the IEEE conference on computer vision and pattern recognition, 2018, pp. 586–595

    2018

  20. [28]

    Image quality assessment: from error visibility to structural similarity,

    Z. Wang, A. C. Bovik, H. R. Sheikh, and E. P. Simoncelli, “Image quality assessment: from error visibility to structural similarity,” IEEE transactions on image processing , vol. 13, no. 4, pp. 600–612, 2004

    2004

  21. [29]

    Underexposed photo enhancement using deep illumination estimation,

    R. Wang, Q. Zhang, C.-W. Fu, X. Shen, W.-S. Zheng, and J. Jia, “Underexposed photo enhancement using deep illumination estimation,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2019, pp. 6849–6857

    2019

  22. [30]

    An underwater image enhancement benchmark dataset and beyond,

    C. Li, C. Guo, W. Ren, R. Cong, J. Hou, S. Kwong, and D. Tao, “An underwater image enhancement benchmark dataset and beyond,” IEEE Transactions on Image Processing , vol. 29, pp. 4376–4389, 2020

    2020

  23. [31]

    Semantic Segmentation of Underwater Imagery: Dataset and Benchmark,

    M. J. Islam, C. Edge, Y . Xiao, P. Luo, M. Mehtaz, C. Morse, S. S. Enan, and J. Sattar, “Semantic Segmentation of Underwater Imagery: Dataset and Benchmark,” in IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) . IEEE/RSJ, 2020

    2020

  24. [32]

    Real-world underwater enhancement: Challenges, benchmarks, and solutions under natural light,

    R. Liu, X. Fan, M. Zhu, M. Hou, and Z. Luo, “Real-world underwater enhancement: Challenges, benchmarks, and solutions under natural light,” IEEE Transactions on Circuits and Systems for Video Technology, vol. 30, no. 12, pp. 4861–4875, 2020

    2020

  25. [33]

    Image quality assessment through fsim, ssim, mse and psnr—a comparative study,

    U. Sara, M. Akter, and M. S. Uddin, “Image quality assessment through fsim, ssim, mse and psnr—a comparative study,” Journal of Computer and Communications, vol. 07, pp. 8–18, 01 2019

    2019

  26. [34]

    Human-visual-system-inspired underwater image quality measures,

    K. Panetta, C. Gao, and S. Agaian, “Human-visual-system-inspired underwater image quality measures,” IEEE Journal of Oceanic Engi- neering, vol. 41, no. 3, pp. 541–551, 2016

    2016

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.