pith. sign in

arxiv: 2606.26295 · v1 · pith:KTT2EOH2new · submitted 2026-06-24 · 💻 cs.CV

Beyond Aesthetics: Quantifying Information Loss in Turbid Scenes

Vasiliki Ismiroglou , Stefan H. Bengtson , Tasos Benos , Thomas B. Moeslund , Malte Pedersen This is my paper

Pith reviewed 2026-06-26 01:27 UTC · model grok-4.3

classification 💻 cs.CV
keywords turbid underwater imagesphase congruencyinformation lossinstance segmentationimage quality metricTUB datasetunderwater computer vision
0
0 comments X

The pith

PCD, derived from phase congruency maps, correlates strongly with instance segmentation performance in turbid underwater scenes while standard metrics do not.

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

Standard image quality measures do not track how computer vision models actually degrade when underwater scenes become turbid. The authors assembled the TUB dataset of 1,320 real extreme-turbidity images carrying more than 16,000 segmentation masks. They introduce PCD, computed from phase congruency maps, as a contrast-invariant indicator of structural information loss. Experiments demonstrate that PCD values align closely with model accuracy on both the real TUB images and synthetic turbid counterparts, whereas conventional metrics exhibit little or no relationship.

Core claim

PCD correlates strongly with the performance of instance segmentation models on both real and synthetic turbid images, whereas common metrics in the field show weak to no correlation at all.

What carries the argument

PCD, a metric derived from phase congruency maps that remains invariant to contrast and quantifies loss of structural information.

If this is right

  • Segmentation accuracy under turbidity can be estimated from PCD alone without executing the models.
  • Synthetic turbidity generation methods can be validated by checking whether their PCD-to-performance curves match those observed on real data.
  • Underwater vision pipelines can be ranked for robustness using PCD as a proxy for expected information loss.
  • Contrast normalization alone is insufficient to restore performance; preserving the structures measured by phase congruency is required.

Where Pith is reading between the lines

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

  • PCD could be used to select or augment training images that preserve structural content rather than merely increasing contrast.
  • Analogous phase-congruency measures might quantify information loss in other scattering environments such as atmospheric haze or medical imaging through tissue.
  • The TUB dataset supplies a benchmark for training models that adapt explicitly to measured levels of structural degradation.

Load-bearing premise

Phase congruency maps computed on real turbid images isolate structural information loss in a manner that stays unchanged by contrast variations and that this loss directly accounts for drops in downstream model accuracy.

What would settle it

A new collection of turbid images in which PCD values fail to predict segmentation accuracy across multiple models would disprove the reported correlation.

Figures

Figures reproduced from arXiv: 2606.26295 by Malte Pedersen, Stefan H. Bengtson, Tasos Benos, Thomas B. Moeslund, Vasiliki Ismiroglou.

Figure 1
Figure 1. Figure 1: Four clear images from the TUB dataset are shown with progressively reduced contrast. The proposed PCD metric indicates little change in information while al￾ternative metrics show a substantial drop in image quality. The corresponding phase congruency maps, computed from the low contrast images, highlight that a consider￾able amount of structural information is retained. As a result, peripheral light ente… view at source ↗
Figure 2
Figure 2. Figure 2: Example of different scenes from the TUB dataset with different levels of tur￾bidity (grouped by NTU) and from different camera-viewpoints. rics are also likely unsuitable for evaluating whether synthetic turbidity effects reproduce the information loss observed in real environments. These findings point to the need for a dedicated evaluation measure for turbid scenes. Developing such a measure requires a … view at source ↗
Figure 3
Figure 3. Figure 3: Annotation pipeline. Masks are delineated in the clear images and extended to all turbidity levels due to the static nature of the setup. 4 Delentropy of Phase Congruency Maps While a vast range of image quality and complexity metrics exist and have been used in the underwater domain, the majority of them aim to align with human perception. However, deep-learning architectures have the theoretical capacity… view at source ↗
Figure 4
Figure 4. Figure 4: The performance reported as AP50 for MaskRCNN, YOLOv11, and Mask2Former when trained and tested on low, medium, and high turbidity images from the TUB dataset. 5.2 PCD Evaluation We additionally aim to evaluate how well existing image- quality and complex￾ity metrics, including our proposed PCD, correlate with performance in down￾stream tasks when using turbid data. For this task we include both real image… view at source ↗
Figure 5
Figure 5. Figure 5: Example of predicted masks from the Mask2Former model trained on the TUB dataset. All images are from the validation split. Colors are for visualization only. mask annotations as their real counterparts, and both sets are merged to form an expanded dataset. In total, this triples the volume of the images. Similar to the turbidity baseline, we train MaskRCNN, YOLOv11, and Mask2- Former on the combined datas… view at source ↗
Figure 6
Figure 6. Figure 6: Metric correlation with model performance. The area highlights do not repre￾sent any measurable quantity, but are simply used to assist visualization. higher scores, similar to the ones of clear images, for the majority of Synth1. While this results in some outliers, there is also a meaningful portion of overlap with the ground-truth distribution. A visual example to explain this behavior can be seen in Fi… view at source ↗
Figure 7
Figure 7. Figure 7: Scatterplots highlighting correlation between model performance and common image quality metrics for real and synthetic data. The models listed from top to bottom are MaskRCNN, YOLOv11, and Mask2Former. NIQE is inverted for visual consistency [PITH_FULL_IMAGE:figures/full_fig_p012_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Example of predicted masks on synthetic images. First row: ground truth masks on non-turbid images from the TUB dataset. Second and third row: predicted masks on images with synthetic turbidity generated using Synth1 and Synth2, respectively. Predictions are from Mask2Former, and all images are from the validation split. Colors are for visualization only. 2.1. However, our analysis indicates that six wavel… view at source ↗
Figure 9
Figure 9. Figure 9: Variation of PCD correlation to Mask2Former AP50 as the wavelet character￾istic over which phase congruency is calculated change [PITH_FULL_IMAGE:figures/full_fig_p013_9.png] view at source ↗
read the original abstract

Visibility in underwater environments degrades rapidly under turbid conditions, yet the effects on computer-vision models remain unclear. This issue is compounded by reliance on synthetic turbidity datasets, which may misrepresent real-world information loss. To address this gap, we introduce the Turbid Underwater Baseline (TUB) dataset, comprising 1,320 images captured under extreme turbidity and over 16,000 high-confidence ground-truth segmentation masks. We additionally propose PCD, a metric derived from phase congruency maps that is invariant to contrast and aims to capture the loss of structural information in real turbidity. We show that PCD correlates strongly with the performance of instance segmentation models on both real and synthetic turbid images, whereas common metrics in the field show weak to no correlation at all. The dataset and relevant code can be found on the project page: https://vap.aau.dk/pcd

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

2 major / 2 minor

Summary. The paper introduces the Turbid Underwater Baseline (TUB) dataset of 1,320 real extreme-turbidity underwater images together with >16,000 high-confidence instance-segmentation masks. It defines PCD, a metric extracted from phase-congruency maps, asserts that PCD is contrast-invariant, and claims that PCD exhibits strong correlation with instance-segmentation model performance on both real and synthetic turbid images while conventional metrics (SSIM, PSNR, etc.) show weak or null correlation.

Significance. If the reported correlations prove robust after proper controls, the work supplies a practical, task-relevant measure of structural information loss under turbidity that could guide underwater vision research beyond purely aesthetic image-quality metrics; the public release of the dataset and code strengthens the contribution.

major comments (2)
  1. [Abstract and §4] Abstract and §4 (Results): the central claim that PCD 'correlates strongly' while other metrics do not is stated without any numerical values, confidence intervals, correlation coefficients, or description of the exact procedure (Pearson/Spearman, number of models, train/test split). This absence prevents verification of the empirical result that is load-bearing for the paper's contribution.
  2. [§3 and §4.2] §3 (PCD definition) and §4.2 (Invariance experiments): the assertion that PCD is invariant to contrast while specifically tracking turbidity-induced structural loss lacks an explicit control experiment that applies pure contrast scaling (no added scattering) to clear images and reports PCD stability. Without this ablation the observed superiority over SSIM/PSNR could be explained by PCD's residual contrast sensitivity rather than unique structural capture.
minor comments (2)
  1. [§2] Clarify in §2 how the 16,000 masks were generated and what 'high-confidence' threshold was applied.
  2. [§4] Add error bars or statistical significance tests to all correlation plots in §4.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive comments that highlight areas where the manuscript can be strengthened for clarity and rigor. We address each major comment below and indicate the revisions we will make.

read point-by-point responses

  1. Referee: [Abstract and §4] Abstract and §4 (Results): the central claim that PCD 'correlates strongly' while other metrics do not is stated without any numerical values, confidence intervals, correlation coefficients, or description of the exact procedure (Pearson/Spearman, number of models, train/test split). This absence prevents verification of the empirical result that is load-bearing for the paper's contribution.

    Authors: We agree that the absence of specific numerical values and procedural details in the abstract and §4 limits verifiability. In the revised manuscript we will report the exact correlation coefficients (Pearson and/or Spearman), associated confidence intervals, the number of models evaluated, the train/test split details, and the precise statistical procedure used to establish the correlations between PCD (and baseline metrics) and instance-segmentation performance. revision: yes

  2. Referee: [§3 and §4.2] §3 (PCD definition) and §4.2 (Invariance experiments): the assertion that PCD is invariant to contrast while specifically tracking turbidity-induced structural loss lacks an explicit control experiment that applies pure contrast scaling (no added scattering) to clear images and reports PCD stability. Without this ablation the observed superiority over SSIM/PSNR could be explained by PCD's residual contrast sensitivity rather than unique structural capture.

    Authors: The referee correctly identifies that an explicit control for pure contrast scaling is missing. We will add a dedicated ablation experiment in the revised §4.2 that applies monotonic contrast scaling (without any scattering) to a set of clear images and reports the resulting PCD values to demonstrate stability, thereby isolating contrast invariance from turbidity-induced structural degradation. revision: yes

Circularity Check

0 steps flagged

No circularity: PCD is an independent empirical metric with no reduction to fitted inputs or self-referential definitions.

full rationale

The paper defines PCD explicitly as a metric computed from phase-congruency maps (an established external technique) and reports its correlation with segmentation performance as an empirical observation on the new TUB dataset. No equations, parameter fits, or self-citations are shown that would make the reported correlation or invariance claim reduce to the input data by construction. The derivation chain is therefore self-contained and externally falsifiable.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract-only review supplies no equations or methods sections, so free parameters, axioms, and invented entities cannot be enumerated. Phase congruency is treated as a standard technique imported from prior literature.

pith-pipeline@v0.9.1-grok · 5693 in / 1170 out tokens · 17927 ms · 2026-06-26T01:27:50.027762+00:00 · methodology

discussion (0)

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

Reference graph

Works this paper leans on

56 extracted references · 41 canonical work pages · 5 internal anchors

  1. [1]

    In: 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition

    Akkaynak, D., Treibitz, T.: A Revised Underwater Image Formation Model. In: 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 6723–6732 (Jun 2018).https://doi.org/10.1109/CVPR.2018.00703,https:// ieeexplore.ieee.org/document/8578801

    work page doi:10.1109/cvpr.2018.00703 2018

  2. [2]

    Deep Underwater Image Enhancement

    Anwar, S., Li, C., Porikli, F.: Deep Underwater Image Enhancement (Jul 2018). https://doi.org/10.48550/arXiv.1807.03528,http://arxiv.org/abs/1807. 03528, arXiv:1807.03528 [cs]

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.1807.03528 2018

  3. [3]

    48550/arXiv.2302.12288,http://arxiv.org/abs/2302.12288, arXiv:2302.12288 [cs]

    Bhat, S.F., Birkl, R., Wofk, D., Wonka, P., Müller, M.: ZoeDepth: Zero-shot Trans- fer by Combining Relative and Metric Depth (Feb 2023).https://doi.org/10. 48550/arXiv.2302.12288,http://arxiv.org/abs/2302.12288, arXiv:2302.12288 [cs]

    Pith/arXiv arXiv 2023

  4. [4]

    Cheng, B., Misra, I., Schwing, A.G., Kirillov, A., Girdhar, R.: Masked-attention Mask Transformer for Universal Image Segmentation (Jun 2022).https : / / doi.org/10.48550/arXiv.2112.01527,http://arxiv.org/abs/2112.01527, arXiv:2112.01527 [cs]

    work page doi:10.48550/arxiv.2112.01527 2022

  5. [5]

    ACM Trans

    Desai, C., Benur, S., Patil, U., Mudenagudi, U.: RSUIGM: Realistic Synthetic Underwater Image Generation with Image Formation Model. ACM Trans. Multi- media Comput. Commun. Appl.21(1), 10:1–10:22 (Dec 2024).https://doi.org/ 10.1145/3656473,https://doi.org/10.1145/3656473

    work page doi:10.1145/3656473 2024

  6. [6]

    In: OCEANS 2016 - Shanghai

    Duarte, A., Codevilla, F., Gaya, J.D.O., Botelho, S.S.C.: A dataset to evaluate underwater image restoration methods. In: OCEANS 2016 - Shanghai. pp. 1– 6 (Apr 2016).https://doi.org/10.1109/OCEANSAP.2016.7485524,https:// ieeexplore.ieee.org/abstract/document/7485524

    work page doi:10.1109/oceansap.2016.7485524 2016

  7. [7]

    In: 2018 IEEE International Conference on Robotics and Automation (ICRA)

    Fabbri, C., Islam, M.J., Sattar, J.: Enhancing Underwater Imagery Using Gener- ative Adversarial Networks. In: 2018 IEEE International Conference on Robotics and Automation (ICRA). pp. 7159–7165 (May 2018).https://doi.org/10.1109/ ICRA.2018.8460552,https://ieeexplore.ieee.org/document/8460552

    arXiv 2018

  8. [8]

    In: 2023 International Conference on Digital Image Computing: Techniques and Applications (DICTA)

    Fathy, M.E., Mohamed, S.A., Awad, M.I., Munim, H.E.A.E.: SubmergeStyleGAN: Synthetic Underwater Data Generation with Style Transfer for Domain Adapta- tion. In: 2023 International Conference on Digital Image Computing: Techniques and Applications (DICTA). pp. 546–553 (Nov 2023).https://doi.org/10.1109/ DICTA60407.2023.00081,https://ieeexplore.ieee.org/doc...

    arXiv 2023

  9. [9]

    Susceptibility of widely diverse influenza A viruses to PB2 polymerase inhibitor pimodivir

    Fu, C., Liu, R., Fan, X., Chen, P., Fu, H., Yuan, W., Zhu, M., Luo, Z.: Re- thinking general underwater object detection: Datasets, challenges, and solu- tions. Neurocomputing517, 243–256 (Jan 2023).https://doi.org/10.1016/j. neucom.2022.10.039,https://www.sciencedirect.com/science/article/pii/ S0925231222013169

    work page doi:10.1016/j 2023

  10. [10]

    ACM Comput

    González-Sabbagh, S.P., Robles-Kelly, A.: A Survey on Underwater Computer Vi- sion. ACM Comput. Surv.55(13s), 268:1–268:39 (Jul 2023).https://doi.org/ 10.1145/3578516,https://dl.acm.org/doi/10.1145/3578516 16 V. Ismiroglou et al

    work page doi:10.1145/3578516 2023

  11. [11]

    IEEE Transactions on Systems, Man, and CyberneticsSMC-3(6), 610–621 (Nov 1973).https://doi.org/10.1109/TSMC.1973.4309314,https: //ieeexplore.ieee.org/document/4309314

    Haralick, R.M., Shanmugam, K., Dinstein, I.: Textural Features for Image Clas- sification. IEEE Transactions on Systems, Man, and CyberneticsSMC-3(6), 610–621 (Nov 1973).https://doi.org/10.1109/TSMC.1973.4309314,https: //ieeexplore.ieee.org/document/4309314

    work page doi:10.1109/tsmc.1973.4309314 1973

  12. [12]

    He, K., Gkioxari, G., Dollár, P., Girshick, R.: Mask R-CNN (Jan 2018).https: //doi.org/10.48550/arXiv.1703.06870,http://arxiv.org/abs/1703.06870, arXiv:1703.06870 [cs]

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.1703.06870 2018

  13. [13]

    He, K., Zhang, X., Ren, S., Sun, J.: Deep Residual Learning for Image Recognition (Dec 2015).https://doi.org/10.48550/arXiv.1512.03385,http://arxiv.org/ abs/1512.03385, arXiv:1512.03385 [cs]

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.1512.03385 2015

  14. [14]

    Horé, A., Ziou, D.: Image Quality Metrics: PSNR vs. SSIM. In: 2010 20th Inter- national Conference on Pattern Recognition. pp. 2366–2369 (Aug 2010).https: //doi.org/10.1109/ICPR.2010.579,https://ieeexplore- ieee- org.zorac. aub.aau.dk/document/5596999, iSSN: 1051-4651

    work page doi:10.1109/icpr.2010.579 2010

  15. [15]

    IEEE Access8, 122078–122091 (2020).https://doi.org/10.1109/ACCESS.2020.3006359,https://ieeexplore

    Hou, G., Zhao, X., Pan, Z., Yang, H., Tan, L., Li, J.: Benchmarking Underwater Image Enhancement and Restoration, and Beyond. IEEE Access8, 122078–122091 (2020).https://doi.org/10.1109/ACCESS.2020.3006359,https://ieeexplore. ieee.org/document/9130676, conference Name: IEEE Access

    work page doi:10.1109/access.2020.3006359 2020

  16. [16]

    Journal of Marine Science and Engineering10(2) (Feb 2022).https://doi.org/10.3390/ jmse10020241,https://www.mdpi.com/2077-1312/10/2/241

    Hu, K., Weng, C., Zhang, Y., Jin, J., Xia, Q.: An Overview of Underwater Vision Enhancement: From Traditional Methods to Recent Deep Learning. Journal of Marine Science and Engineering10(2) (Feb 2022).https://doi.org/10.3390/ jmse10020241,https://www.mdpi.com/2077-1312/10/2/241

    2022

  17. [17]

    IEEE Robotics and Automation Letters5(2), 3227–3234 (Apr 2020).https://doi.org/10.1109/LRA.2020.2974710,https://ieeexplore

    Islam, M.J., Xia, Y., Sattar, J.: Fast Underwater Image Enhancement for Improved Visual Perception. IEEE Robotics and Automation Letters5(2), 3227–3234 (Apr 2020).https://doi.org/10.1109/LRA.2020.2974710,https://ieeexplore. ieee.org/document/9001231

    work page doi:10.1109/lra.2020.2974710 2020

  18. [18]

    In: In Proceedings of the IEEE/CVF International Conference on Computer Vision Workshops (2025)

    Ismiroglou, V., Pedersen, M., Bengtson, S.H., Aakerberg, A., Moeslund, T.B.: Sea- ing Through Scattered Rays: Revisiting the Image Formation Model for Realistic Underwater Image Generation. In: In Proceedings of the IEEE/CVF International Conference on Computer Vision Workshops (2025)

    2025

  19. [19]

    IEEE Journal of Oceanic Engineering15(2), 101–111 (Apr 1990).https: //doi.org/10.1109/48.50695,https://ieeexplore-ieee-org.zorac.aub.aau

    Jaffe, J.: Computer modeling and the design of optimal underwater imaging sys- tems. IEEE Journal of Oceanic Engineering15(2), 101–111 (Apr 1990).https: //doi.org/10.1109/48.50695,https://ieeexplore-ieee-org.zorac.aub.aau. dk/document/50695

    work page doi:10.1109/48.50695 1990

  20. [20]

    Ecological Informatics78, 102303 (Dec 2023).https://doi.org/ 10.1016/j.ecoinf.2023.102303,https://www.sciencedirect.com/science/ article/pii/S1574954123003321

    Jahanbakht,M.,RahimiAzghadi,M.,Waltham,N.J.:Semi-supervisedandweakly- supervised deep neural networks and dataset for fish detection in turbid under- water videos. Ecological Informatics78, 102303 (Dec 2023).https://doi.org/ 10.1016/j.ecoinf.2023.102303,https://www.sciencedirect.com/science/ article/pii/S1574954123003321

    work page doi:10.1016/j.ecoinf.2023.102303 2023

  21. [21]

    Computer-Aided Civil and Infrastructure Engineer- ing39(2), 281–299 (2024).https : / / doi

    Jiao, P., Ye, X., Zhang, C., Li, W., Wang, H.: Vision-based real-time marine and offshore structural health monitoring system using un- derwater robots. Computer-Aided Civil and Infrastructure Engineer- ing39(2), 281–299 (2024).https : / / doi . org / 10 . 1111 / mice . 12993, https://onlinelibrary.wiley.com/doi/abs/10.1111/mice.12993, _eprint: https://on...

    work page doi:10.1111/mice.12993 2024

  22. [22]

    com/ultralytics/ultralytics, original-date: 2022-09-11T16:39:45Z

    Jocher, G., Qiu, J., Chaurasia, A.: Ultralytics YOLO (Jan 2023),https://github. com/ultralytics/ultralytics, original-date: 2022-09-11T16:39:45Z

    2023

  23. [23]

    Psychological Research 64(2), 136–148 (Dec 2000).https://doi.org/10.1007/s004260000024,https: //doi.org/10.1007/s004260000024 PCD 17

    Kovesi, P.: Phase congruency: A low-level image invariant. Psychological Research 64(2), 136–148 (Dec 2000).https://doi.org/10.1007/s004260000024,https: //doi.org/10.1007/s004260000024 PCD 17

    work page doi:10.1007/s004260000024 2000

  24. [24]

    Reflections on Shannon Information: In search of a natural information-entropy for images

    Larkin, K.G.: Reflections on Shannon Information: In search of a natural information-entropy for images (Sep 2016).https://doi.org/10.48550/arXiv. 1609.01117,http://arxiv.org/abs/1609.01117, arXiv:1609.01117 [cs]

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv 2016

  25. [25]

    Li, C., Guo, C., Ren, W., Cong, R., Hou, J., Kwong, S., Tao, D.: An Underwater ImageEnhancementBenchmarkDatasetandBeyond.IEEETransactionsonImage Processing29, 4376–4389 (2020).https://doi.org/10.1109/TIP.2019.2955241, https://ieeexplore.ieee.org/document/8917818

    work page doi:10.1109/tip.2019.2955241 2020

  26. [26]

    IEEE Access6, 54241–54257 (2018).https://doi.org/10.1109/ACCESS.2018

    Li, N., Zheng, Z., Zhang, S., Yu, Z., Zheng, H., Zheng, B.: The Synthesis of Un- paired Underwater Images Using a Multistyle Generative Adversarial Network. IEEE Access6, 54241–54257 (2018).https://doi.org/10.1109/ACCESS.2018. 2870854,https://ieeexplore.ieee.org/abstract/document/8467323

    work page doi:10.1109/access.2018 2018

  27. [27]

    IEEE TransactionsonCircuitsandSystemsforVideoTechnology32(5),2831–2844(May 2022).https://doi.org/10.1109/TCSVT.2021.3100059,http://arxiv.org/abs/ 2003.01446, arXiv:2003.01446 [cs]

    Liu, C., Wang, Z., Wang, S., Tang, T., Tao, Y., Yang, C., Li, H., Liu, X., Fan, X.: A New Dataset, Poisson GAN and AquaNet for Underwater Object Grabbing. IEEE TransactionsonCircuitsandSystemsforVideoTechnology32(5),2831–2844(May 2022).https://doi.org/10.1109/TCSVT.2021.3100059,http://arxiv.org/abs/ 2003.01446, arXiv:2003.01446 [cs]

    work page doi:10.1109/tcsvt.2021.3100059 2022

  28. [28]

    Liu, L., Liu, B., Huang, H., Bovik, A.C.: No-reference image quality assessment based on spatial and spectral entropies. Signal Processing: Image Communication 29(8), 856–863 (Sep 2014).https://doi.org/10.1016/j.image.2014.06.006, http://pure.bit.edu.cn/en/publications/no- reference- image- quality- assessment-based-on-spatial-and-spectr/

    work page doi:10.1016/j.image.2014.06.006 2014

  29. [29]

    Liu, R., Fan, X., Zhu, M., Hou, M., Luo, Z.: Real-world Underwater Enhancement: Challenges, Benchmarks, and Solutions (Mar 2019).https://doi.org/10.48550/ arXiv.1901.05320,http://arxiv.org/abs/1901.05320, arXiv:1901.05320 [cs]

    Pith/arXiv arXiv 2019

  30. [30]

    Information13(4), 187 (Apr 2022).https://doi.org/10.3390/ info13040187,https://www.mdpi.com/2078-2489/13/4/187

    Liu, Y., Xu, H., Zhang, B., Sun, K., Yang, J., Li, B., Li, C., Quan, X.: Model-Based Underwater Image Simulation and Learning-Based Underwater Image Enhance- ment Method. Information13(4), 187 (Apr 2022).https://doi.org/10.3390/ info13040187,https://www.mdpi.com/2078-2489/13/4/187

    2022

  31. [31]

    Swin Transformer: Hierarchical Vision Transformer using Shifted Windows

    Liu, Z., Lin, Y., Cao, Y., Hu, H., Wei, Y., Zhang, Z., Lin, S., Guo, B.: Swin Transformer: Hierarchical Vision Transformer using Shifted Windows (Aug 2021). https://doi.org/10.48550/arXiv.2103.14030,http://arxiv.org/abs/2103. 14030, arXiv:2103.14030 [cs]

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2103.14030 2021

  32. [32]

    Lucas, E., Awad, A., Geglio, A., Moradi, S., Saleem, A., Havens, T., Galloway, A., Paheding, S.: Underwater Image Enhancement and Object Detection: Are Poor Object Detection Results On Enhanced Images Due to Missing Human Labels? In: In Proceedings of the Winter Conference on Applications of Computer Vision Workshops (2025)

    2025

  33. [33]

    In: Ocean Optics VI

    McGlamery, B.L.: A Computer Model For Underwater Camera Systems. In: Ocean Optics VI. vol. 0208, pp. 221–231. SPIE (Mar 1980).https://doi.org/10.1117/ 12.958279,https://www.spiedigitallibrary.org/conference- proceedings- of-spie/0208/0000/A-Computer-Model-For-Underwater-Camera-Systems/10. 1117/12.958279.full

    1980

  34. [34]

    IEEE Transactions on Image Processing21(12), 4695–4708 (Dec 2012).https://doi.org/10.1109/TIP.2012.2214050,http://ieeexplore

    Mittal, A., Moorthy, A.K., Bovik, A.C.: No-Reference Image Quality Assessment in the Spatial Domain. IEEE Transactions on Image Processing21(12), 4695–4708 (Dec 2012).https://doi.org/10.1109/TIP.2012.2214050,http://ieeexplore. ieee.org/document/6272356/

    work page doi:10.1109/tip.2012.2214050 2012

  35. [35]

    Completely Blind

    Mittal, A., Soundararajan, R., Bovik, A.C.: Making a “Completely Blind” Image Quality Analyzer. IEEE Signal Processing Letters20(3), 209–212 (Mar 2013). https://doi.org/10.1109/LSP.2012.2227726,http://ieeexplore.ieee.org/ document/6353522/ 18 V. Ismiroglou et al

    work page doi:10.1109/lsp.2012.2227726 2013

  36. [36]

    Academic (Jan 1994), journal Abbreviation: Academic Press Publication Title: Academic Press

    Mobley, C.: Light and Water: Radiative Transfer in Natural Waters. Academic (Jan 1994), journal Abbreviation: Academic Press Publication Title: Academic Press

    1994

  37. [37]

    IEEE Transactions on Image Processing20(12), 3350–3364 (Dec 2011).https://doi.org/10.1109/TIP.2011.2147325,https: //ieeexplore.ieee.org/document/5756237/

    Moorthy, A.K., Bovik, A.C.: Blind Image Quality Assessment: From Natural Scene Statistics to Perceptual Quality. IEEE Transactions on Image Processing20(12), 3350–3364 (Dec 2011).https://doi.org/10.1109/TIP.2011.2147325,https: //ieeexplore.ieee.org/document/5756237/

    work page doi:10.1109/tip.2011.2147325 2011

  38. [38]

    Nature324, 250–253 (Nov 1986).https://doi.org/10.1038/324250a0, https://ui.adsabs.harvard.edu/abs/1986Natur.324..250M, aDS Bibcode: 1986Natur.324..250M

    Morrone, M.C., Ross, J., Burr, D.C., Owens, R.: Mach bands are phase depen- dent. Nature324, 250–253 (Nov 1986).https://doi.org/10.1038/324250a0, https://ui.adsabs.harvard.edu/abs/1986Natur.324..250M, aDS Bibcode: 1986Natur.324..250M

    work page doi:10.1038/324250a0 1986

  39. [39]

    Multimedia Tools and Applications84(22), 24839–24858 (Jul 2025).https : / / doi

    Naveen, P.: Advancements in underwater imaging through machine learning: Tech- niques, challenges, and applications. Multimedia Tools and Applications84(22), 24839–24858 (Jul 2025).https : / / doi . org / 10 . 1007 / s11042 - 024 - 20091 - 4, https://doi.org/10.1007/s11042-024-20091-4

    work page doi:10.1007/s11042-024-20091-4 2025

  40. [40]

    Proceedings of the IEEE69(5), 529–541 (May 1981).https://doi.org/10.1109/PROC.1981.12022, https://ieeexplore.ieee.org/document/1456290

    Oppenheim, A., Lim, J.: The importance of phase in signals. Proceedings of the IEEE69(5), 529–541 (May 1981).https://doi.org/10.1109/PROC.1981.12022, https://ieeexplore.ieee.org/document/1456290

    work page doi:10.1109/proc.1981.12022 1981

  41. [41]

    IEEE Journal of Oceanic Engineering41(3), 541–551 (Jul 2016).https://doi.org/10.1109/JOE.2015.2469915,https://ieeexplore- ieee-org.zorac.aub.aau.dk/document/7305804

    Panetta, K., Gao, C., Agaian, S.: Human-Visual-System-Inspired Underwater Im- age Quality Measures. IEEE Journal of Oceanic Engineering41(3), 541–551 (Jul 2016).https://doi.org/10.1109/JOE.2015.2469915,https://ieeexplore- ieee-org.zorac.aub.aau.dk/document/7305804

    work page doi:10.1109/joe.2015.2469915 2016

  42. [42]

    In: Proceed- ings of the IEEE/CVF Conference on Computer Vision and Pattern Recogni- tion (CVPR) Workshops

    Pedersen, M., Bruslund Haurum, J., Gade, R., Moeslund, T.B.: Detection of Ma- rine Animals in a New Underwater Dataset with Varying Visibility. In: Proceed- ings of the IEEE/CVF Conference on Computer Vision and Pattern Recogni- tion (CVPR) Workshops. pp. 18–26 (2019),https://openaccess.thecvf.com/ content_CVPRW_2019/html/AAMVEM/Pedersen_Detection_of_Mari...

    2019

  43. [43]

    In: 2020 International Conference on Artificial Intelligence in In- formation and Communication (ICAIIC)

    Rahane, A.A., Subramanian, A.: Measures of Complexity for Large Scale Im- age Datasets. In: 2020 International Conference on Artificial Intelligence in In- formation and Communication (ICAIIC). pp. 282–287 (Feb 2020).https://doi. org/10.1109/ICAIIC48513.2020.9065274,http://arxiv.org/abs/2008.04431, arXiv:2008.04431 [cs]

    work page doi:10.1109/icaiic48513.2020.9065274 2020

  44. [44]

    Artificial Intel- ligence Review54(7), 5413–5467 (Oct 2021).https://doi.org/10.1007/s10462- 021-10025-z,https://doi.org/10.1007/s10462-021-10025-z

    Raveendran, S., Patil, M.D., Birajdar, G.K.: Underwater image enhancement: a comprehensive review, recent trends, challenges and applications. Artificial Intel- ligence Review54(7), 5413–5467 (Oct 2021).https://doi.org/10.1007/s10462- 021-10025-z,https://doi.org/10.1007/s10462-021-10025-z

    work page doi:10.1007/s10462- 2021

  45. [45]

    IEEE Signal Processing Letters17(6), 583–586 (Jun 2010)

    Saad, M.A., Bovik, A.C., Charrier, C.: A DCT Statistics-Based Blind Im- age Quality Index. IEEE Signal Processing Letters17(6), 583–586 (Jun 2010). https://doi.org/10.1109/LSP.2010.2045550,https://ieeexplore.ieee.org/ document/5430991/

    work page doi:10.1109/lsp.2010.2045550 2010

  46. [46]

    IEEE Transactions on Image Processing15(2), 430–444 (Feb 2006).https://doi.org/10.1109/TIP

    Sheikh, H., Bovik, A.: Image information and visual quality. IEEE Transactions on Image Processing15(2), 430–444 (Feb 2006).https://doi.org/10.1109/TIP. 2005.859378,https://ieeexplore.ieee.org/document/1576816/

    work page doi:10.1109/tip 2006

  47. [47]

    Applied Optics54(17), 5392–5401 (Jun 2015).https://doi.org/10.1364/AO.54

    Solonenko, M.G., Mobley, C.D.: Inherent optical properties of Jerlov water types. Applied Optics54(17), 5392–5401 (Jun 2015).https://doi.org/10.1364/AO.54. 005392,https://opg.optica.org/ao/abstract.cfm?uri=ao-54-17-5392

    work page doi:10.1364/ao.54 2015

  48. [48]

    Advanced Engineering Informatics62, 102723 (Oct 2024).https://doi.org/10.1016/j.aei.2024.102723,https://linkinghub

    Sun, W., Hou, S., Wu, G., Zhang, J., Zhang, Y., Jiang, H.: Turbid image tack- ling framework towards underwater concrete bridge detection based on distance control and deep learning. Advanced Engineering Informatics62, 102723 (Oct 2024).https://doi.org/10.1016/j.aei.2024.102723,https://linkinghub. elsevier.com/retrieve/pii/S1474034624003719 PCD 19

    work page doi:10.1016/j.aei.2024.102723 2024

  49. [49]

    In: 2018 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)

    Álvarez Tuñón, O., Rodríguez, A., Jardón, A., Balaguer, C.: Underwater Robot Navigation for Maintenance and Inspection of Flooded Mine Shafts. In: 2018 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS). pp. 1482–1487 (Oct 2018).https : / / doi . org / 10 . 1109 / IROS . 2018 . 8594445, https://ieeexplore.ieee.org/document/8594445/...

    arXiv 2018

  50. [50]

    In: 2019 IEEE International Conference on Image Processing (ICIP)

    Ueda, T., Yamada, K., Tanaka, Y.: Underwater Image Synthesis from RGB-D Im- ages and its Application to Deep Underwater Image Restoration. In: 2019 IEEE International Conference on Image Processing (ICIP). pp. 2115–2119 (Sep 2019). https://doi.org/10.1109/ICIP.2019.8803195,https://ieeexplore.ieee. org/document/8803195/

    work page doi:10.1109/icip.2019.8803195 2019

  51. [51]

    ArXiv (Sep 2019),https : / / www

    Wang, N., Zhou, Y., Han, F., Zhu, H., Zheng, Y.: UWGAN: Un- derwater GAN for Real-world Underwater Color Restoration and De- hazing. ArXiv (Sep 2019),https : / / www . semanticscholar . org / paper / ad8fbe003c65c15429ffa1a81fbd311f9bbae777

    2019

  52. [52]

    Wang, Y., Guo, J., He, W., Gao, H., Yue, H., Zhang, Z., Li, C.: Is Underwater Im- age Enhancement All Object Detectors Need? IEEE Journal of Oceanic Engineer- ing49(2), 606–621 (Apr 2024).https://doi.org/10.1109/JOE.2023.3302888, https://ieeexplore.ieee.org/document/10376393/

    work page doi:10.1109/joe.2023.3302888 2024

  53. [53]

    Wu, Y., Kirillov, A., Massa, F., Lo, W.Y., Girshick, R.: Detectron2.https:// github.com/facebookresearch/detectron2(2019)

    2019

  54. [54]

    Neurocomputing 527, 204–232 (Mar 2023).https://doi.org/10.1016/j.neucom.2023.01.056, https://linkinghub.elsevier.com/retrieve/pii/S0925231223000656

    Xu, S., Zhang, M., Song, W., Mei, H., He, Q., Liotta, A.: A systematic review and analysis of deep learning-based underwater object detection. Neurocomputing 527, 204–232 (Mar 2023).https://doi.org/10.1016/j.neucom.2023.01.056, https://linkinghub.elsevier.com/retrieve/pii/S0925231223000656

    work page doi:10.1016/j.neucom.2023.01.056 2023

  55. [55]

    IEEE Transactions on Image Processing24(12), 6062–6071 (Dec 2015)

    Yang, M., Sowmya, A.: An Underwater Color Image Quality Evaluation Met- ric. IEEE Transactions on Image Processing24(12), 6062–6071 (Dec 2015). https://doi.org/10.1109/TIP.2015.2491020,https://ieeexplore.ieee.org/ document/7300447

    work page doi:10.1109/tip.2015.2491020 2015

  56. [56]

    Applied Sciences12(10) (May 2022).https://doi.org/10.3390/app12104898,https://www.mdpi.com/2076- 3417/12/10/4898

    Yuan, X., Guo, L., Luo, C., Zhou, X., Yu, C.: A Survey of Target Detection and Recognition Methods in Underwater Turbid Areas. Applied Sciences12(10) (May 2022).https://doi.org/10.3390/app12104898,https://www.mdpi.com/2076- 3417/12/10/4898

    work page doi:10.3390/app12104898 2022