pith. machine review for the scientific record. sign in

arxiv: 2512.05683 · v2 · submitted 2025-12-05 · 💻 cs.CV · physics.optics

Recognition: 2 theorem links

· Lean Theorem

Physics-Informed Graph Neural Networks for Frequency-Aware Optical Aberration Correction

Yong En Kok , Bowen Deng , Alexander Bentley , Andrew J. Parkes , Michael G. Somekh , Amanda J. Wright , Michael P. Pound
Authors on Pith no claims yet

Pith reviewed 2026-05-17 00:40 UTC · model grok-4.3

classification 💻 cs.CV physics.optics
keywords optical aberration correctionZernike polynomialsgraph neural networksimage restorationmicroscopyphysics-informed neural networksfrequency domain alignment
0
0 comments X

The pith

A graph neural network models relationships between Zernike polynomials by azimuthal degree to correct large optical aberrations while aligning predictions with frequency-domain features.

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

The paper introduces ZRNet to jointly predict Zernike coefficients and restore images degraded by wavefront distortions in microscopy. It adds a Zernike Graph module that connects polynomial terms according to their azimuthal orders so corrections respect known optical interactions. A Frequency-Aware Alignment loss further requires that coefficient estimates and restored image features agree in the Fourier domain. Experiments on CytoImageNet and real microscope data show gains in both restoration quality and coefficient accuracy for complex, high-amplitude aberrations across sample types and modalities.

Core claim

ZRNet jointly performs Zernike coefficient prediction and optical image restoration. The Zernike Graph module explicitly models physical relationships between Zernike polynomials based on their azimuthal degrees to ensure corrections align with fundamental optical principles. The Frequency-Aware Alignment loss enforces physical consistency by aligning Zernike predictions with image features in the Fourier domain.

What carries the argument

Zernike Graph module that connects Zernike polynomial terms according to azimuthal degrees, paired with Frequency-Aware Alignment loss that matches predictions to Fourier-domain image features.

If this is right

  • The method achieves state-of-the-art results on both image restoration and Zernike coefficient prediction across diverse microscopy modalities and biological samples.
  • Performance holds on experimental point-spread-function data collected from a physical microscope.
  • The network remains effective under realistic sensor noise and generalizes past purely simulated conditions.

Where Pith is reading between the lines

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

  • The same graph construction could be tested on other wavefront correction tasks such as adaptive optics in astronomy.
  • Real-time versions might be paired with deformable mirrors for live-sample correction during imaging sessions.
  • The frequency-alignment idea could extend to related inverse problems where coefficient estimates must stay consistent with measured intensities.

Load-bearing premise

That building a graph from azimuthal degrees between Zernike terms and adding the frequency alignment loss will force the network to follow optical physics rather than simply fitting training examples.

What would settle it

An ablation study on a dataset with known physical aberration patterns in which removing the graph module or the Frequency-Aware Alignment loss produces no measurable drop in restoration or coefficient accuracy would falsify the claim.

Figures

Figures reproduced from arXiv: 2512.05683 by Alexander Bentley, Amanda J. Wright, Andrew J. Parkes, Bowen Deng, Michael G. Somekh, Michael P. Pound, Yong En Kok.

Figure 1
Figure 1. Figure 1: Comparison of aberration severity and correction performance. The top row shows aberrated images: [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 3
Figure 3. Figure 3: Our Zernike graph architecture for processing Zernike modes based on their azimuthal degrees. The [PITH_FULL_IMAGE:figures/full_fig_p015_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Qualitative comparison of SOTA image restoration networks on CytoImageNet [ [PITH_FULL_IMAGE:figures/full_fig_p021_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Zernike coefficients prediction comparing ground truth to ZRNet predictions. 22 [PITH_FULL_IMAGE:figures/full_fig_p023_5.png] view at source ↗
read the original abstract

Optical aberrations significantly degrade image quality in microscopy, particularly when imaging deeper into samples. These aberrations arise from distortions in the optical wavefront and can be mathematically represented using Zernike polynomials. Existing methods often address only mild aberrations on limited sample types and modalities, typically treating the problem as a black-box mapping without leveraging the underlying optical physics of wavefront distortions. We propose ZRNet, a physics-informed framework that jointly performs Zernike coefficient prediction and optical image Restoration. We contribute a Zernike Graph module that explicitly models physical relationships between Zernike polynomials based on their azimuthal degrees-ensuring that learned corrections align with fundamental optical principles. To further enforce physical consistency between image restoration and Zernike prediction, we introduce a Frequency-Aware Alignment (FAA) loss, which better aligns Zernike coefficient prediction and image features in the Fourier domain. Extensive experiments on CytoImageNet demonstrates that our approach achieves state-of-the-art performance in both image restoration and Zernike coefficient prediction across diverse microscopy modalities and biological samples with complex, large-amplitude aberrations. We further validate on experimental PSF data from a physical microscope and demonstrate robustness to realistic sensor noise, confirming generalisation beyond simulated conditions. Code is available at https://github.com/janetkok/ZRNet.

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 ZRNet, a physics-informed graph neural network for joint Zernike coefficient prediction and optical image restoration in microscopy. It contributes a Zernike Graph module that connects polynomials according to azimuthal degree m to enforce optical relationships, plus a Frequency-Aware Alignment (FAA) loss that aligns features in the Fourier domain. The central claim is state-of-the-art performance on CytoImageNet for both restoration and coefficient prediction across modalities and large-amplitude aberrations, with additional validation on real experimental PSF data and robustness to sensor noise.

Significance. If the performance edge is shown to arise from the physics-informed components rather than added capacity, the work would provide a useful template for embedding wavefront physics into learned aberration correctors, with direct relevance to deep-tissue microscopy. The release of code and the inclusion of real PSF validation are concrete strengths that support reproducibility and generalization claims.

major comments (2)
  1. [Method and Experiments] The abstract and method description assert that the Zernike Graph (azimuthal-degree edges) plus FAA loss produce corrections that 'align with fundamental optical principles.' No ablation replaces the graph with a standard message-passing layer of matched capacity, nor is wavefront error reported on held-out experimental PSFs; without these controls the physics-informed framing remains untested against the alternative that the modules simply increase expressivity on the simulated training distribution.
  2. [Experiments] Table and figure captions reference SOTA results on CytoImageNet, yet the manuscript does not report the exact data splits, the full set of baselines (including non-graph CNNs and physics-agnostic GNNs), or statistical significance of the reported gains; these omissions make it impossible to assess whether the claimed superiority is robust or sensitive to post-hoc choices.
minor comments (2)
  1. [Method] Notation for the Zernike Graph adjacency matrix should be introduced once with a clear equation rather than described only in prose.
  2. [Method] The FAA loss is defined in the Fourier domain; a short derivation or reference showing why this particular alignment term is preferred over direct coefficient MSE would improve clarity.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the thoughtful and constructive comments. We address each major point below and describe the revisions that will be incorporated into the next version of the manuscript.

read point-by-point responses

  1. Referee: [Method and Experiments] The abstract and method description assert that the Zernike Graph (azimuthal-degree edges) plus FAA loss produce corrections that 'align with fundamental optical principles.' No ablation replaces the graph with a standard message-passing layer of matched capacity, nor is wavefront error reported on held-out experimental PSFs; without these controls the physics-informed framing remains untested against the alternative that the modules simply increase expressivity on the simulated training distribution.

    Authors: We agree that isolating the contribution of the physics-informed design is important. In the revised manuscript we will add an ablation that replaces the Zernike Graph module with a standard message-passing GNN layer of matched capacity and report the resulting performance on both coefficient prediction and image restoration. We will also report wavefront error (Zernike coefficient RMSE) on the held-out experimental PSF data to provide direct evidence that the learned corrections generalize beyond the simulated distribution. revision: yes

  2. Referee: [Experiments] Table and figure captions reference SOTA results on CytoImageNet, yet the manuscript does not report the exact data splits, the full set of baselines (including non-graph CNNs and physics-agnostic GNNs), or statistical significance of the reported gains; these omissions make it impossible to assess whether the claimed superiority is robust or sensitive to post-hoc choices.

    Authors: We acknowledge these omissions limit the ability to fully evaluate the claims. The revised manuscript will explicitly document the train/validation/test splits used for CytoImageNet, expand the baseline set to include non-graph CNNs and physics-agnostic GNN variants, and report statistical significance (e.g., paired statistical tests with p-values) for the observed performance differences. revision: yes

Circularity Check

0 steps flagged

No circularity: architecture and loss are independent design choices validated on external data

full rationale

The paper defines the Zernike Graph from the known azimuthal order m of Zernike polynomials and introduces the FAA loss as a Fourier-domain alignment term; neither step reduces any claimed prediction or coefficient to a fitted input by construction. The central claims rest on end-to-end training plus experimental validation on CytoImageNet and real PSF measurements, not on self-referential equations or load-bearing self-citations. No derivation chain collapses to tautology.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The framework rests on standard optical assumptions about Zernike representation and introduces no new free parameters or invented physical entities beyond standard neural-network hyperparameters.

axioms (1)
  • domain assumption Zernike polynomials provide a complete and accurate basis for representing optical aberrations in microscopy
    Invoked in the problem formulation and method design.

pith-pipeline@v0.9.0 · 5547 in / 1238 out tokens · 63096 ms · 2026-05-17T00:40:10.688011+00:00 · methodology

discussion (0)

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

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

69 extracted references · 69 canonical work pages · 1 internal anchor

  1. [1]

    K. M. Hampson, R. Turcotte, D. T. Miller, K. Kurokawa, J. R. Males, N. Ji, M. J. Booth, Adaptive optics for high-resolution imaging, Nature Reviews Methods Primers 1 (1) (2021) 68

    work page 2021

  2. [2]

    Schwertner, M

    M. Schwertner, M. J. Booth, M. A. Neil, T. Wilson, Measurement of specimen- induced aberrations of biological samples using phase stepping interferometry, Journal of microscopy 213 (1) (2004) 11–19

    work page 2004

  3. [3]

    J. W. Hardy, Adaptive optics for astronomical telescopes, V ol. 16, Oxford Univer- sity Press on Demand, 1998

    work page 1998

  4. [4]

    B. C. Platt, R. Shack, History and principles of shack-hartmann wavefront sensing (2001)

    work page 2001

  5. [5]

    X. Tao, B. Fernandez, O. Azucena, M. Fu, D. Garcia, Y . Zuo, D. C. Chen, J. Kubby, Adaptive optics confocal microscopy using direct wavefront sensing, Optics letters 36 (7) (2011) 1062–1064

    work page 2011

  6. [6]

    M. A. V orontsov, G. W. Carhart, M. Cohen, G. Cauwenberghs, Adaptive op- tics based on analog parallel stochastic optimization: analysis and experimental demonstration, JOSA A 17 (8) (2000) 1440–1453

    work page 2000

  7. [7]

    Zommer, E

    S. Zommer, E. Ribak, S. Lipson, J. Adler, Simulated annealing in ocular adaptive optics, Optics letters 31 (7) (2006) 939–941. 27

    work page 2006

  8. [8]

    P. Yang, M. Ao, Y . Liu, B. Xu, W. Jiang, Intracavity transverse modes controlled by a genetic algorithm based on zernike mode coefficients, Optics express 15 (25) (2007) 17051–17062

    work page 2007

  9. [9]

    R. W. Gerchberg, A practical algorithm for the determination of plane from image and diffraction pictures, Optik 35 (2) (1972) 237–246

    work page 1972

  10. [10]

    A. J. Janssen, Extended nijboer–zernike approach for the computation of optical point-spread functions, JOSA A 19 (5) (2002) 849–857

    work page 2002

  11. [11]

    J. Liu, P. Wang, X. Zhang, Y . He, X. Zhou, H. Ye, Y . Li, S. Xu, S. Chen, D. Fan, Deep learning based atmospheric turbulence compensation for orbital angular momentum beam distortion and communication, Optics express 27 (12) (2019) 16671–16688

    work page 2019

  12. [12]

    H. Guo, Y . Xu, Q. Li, S. Du, D. He, Q. Wang, Y . Huang, Improved machine learning approach for wavefront sensing, Sensors 19 (16) (2019) 3533

    work page 2019

  13. [13]

    K. Wang, M. Zhang, J. Tang, L. Wang, L. Hu, X. Wu, W. Li, J. Di, G. Liu, J. Zhao, Deep learning wavefront sensing and aberration correction in atmospheric turbulence, PhotoniX 2 (1) (2021) 1–11

    work page 2021

  14. [14]

    Nishizaki, M

    Y . Nishizaki, M. Valdivia, R. Horisaki, K. Kitaguchi, M. Saito, J. Tanida, E. Vera, Deep learning wavefront sensing, Optics express 27 (1) (2019) 240–251

    work page 2019

  15. [15]

    Y . Xu, D. He, Q. Wang, H. Guo, Q. Li, Z. Xie, Y . Huang, An improved method of measuring wavefront aberration based on image with machine learning in free space optical communication, Sensors 19 (17) (2019) 3665

    work page 2019

  16. [16]

    Q. Xin, G. Ju, C. Zhang, S. Xu, Object-independent image-based wavefront sensing approach using phase diversity images and deep learning, Optics express 27 (18) (2019) 26102–26119

    work page 2019

  17. [17]

    Q. Hu, M. Hailstone, J. Wang, M. Wincott, D. Stoychev, H. Atilgan, D. Gala, T. Chaiamarit, R. M. Parton, J. Antonello, et al., Universal adaptive op- tics for microscopy through embedded neural network control, arXiv preprint arXiv:2301.02647 (2023). 28

    work page arXiv 2023

  18. [18]

    M. R. Rai, C. Li, H. T. Ghashghaei, A. Greenbaum, Deep learning-based adaptive optics for light sheet fluorescence microscopy, Biomedical Optics Express 14 (6) (2023) 2905–2919

    work page 2023

  19. [19]

    D. Fish, A. Brinicombe, E. Pike, J. Walker, Blind deconvolution by means of the richardson–lucy algorithm, JOSA A 12 (1) (1995) 58–65

    work page 1995

  20. [20]

    Shajkofci, M

    A. Shajkofci, M. Liebling, Semi-blind spatially-variant deconvolution in optical microscopy with local point spread function estimation by use of convolutional neu- ral networks, in: 2018 25th IEEE International Conference on Image Processing (ICIP), IEEE, 2018, pp. 3818–3822

    work page 2018

  21. [21]

    R. K. Tyson, B. W. Frazier, Principles of adaptive optics, CRC press, 2022

    work page 2022

  22. [22]

    M. Guo, Y . Wu, C. M. Hobson, Y . Su, S. Qian, E. Krueger, R. Christensen, G. Kroeschell, J. Bui, M. Chaw, et al., Deep learning-based aberration com- pensation improves contrast and resolution in fluorescence microscopy, Nature Communications 16 (1) (2025) 313

    work page 2025

  23. [23]

    A. P. Krishnan, C. Belthangady, C. Nyby, M. Lange, B. Yang, L. A. Royer, Optical aberration correction via phase diversity and deep learning, BioRxiv (2020) 2020– 04

    work page 2020

  24. [24]

    I. Kang, Q. Zhang, S. X. Yu, N. Ji, Coordinate-based neural representa- tions for computational adaptive optics in widefield microscopy, arXiv preprint arXiv:2307.03812 (2023)

    work page arXiv 2023

  25. [25]

    C. Qiao, H. Chen, R. Wang, T. Jiang, Y . Wang, D. Li, Deep learning-based optical aberration estimation enables offline digital adaptive optics and super-resolution imaging, Photonics Research 12 (3) (2024) 474–484

    work page 2024

  26. [26]

    Schwertner, M

    M. Schwertner, M. Booth, T. Wilson, Simulation of specimen-induced aberrations for objects with spherical and cylindrical symmetry, Journal of microscopy 215 (3) (2004) 271–280. 29

    work page 2004

  27. [27]

    K. Wang, W. Sun, C. T. Richie, B. K. Harvey, E. Betzig, N. Ji, Direct wavefront sensing for high-resolution in vivo imaging in scattering tissue, Nature communi- cations 6 (1) (2015) 7276

    work page 2015

  28. [28]

    M. J. Booth, Adaptive optical microscopy: the ongoing quest for a perfect image, Light: Science & Applications 3 (4) (2014) e165–e165

    work page 2014

  29. [29]

    S. B. Z. Hua, A. X. Lu, A. M. Moses, Cytoimagenet: A large-scale pretraining dataset for bioimage transfer learning, arXiv preprint arXiv:2111.11646 (2021)

    work page arXiv 2021

  30. [30]

    Liang, J

    J. Liang, J. Cao, G. Sun, K. Zhang, L. Van Gool, R. Timofte, Swinir: Image restoration using swin transformer, in: Proceedings of the IEEE/CVF international conference on computer vision, 2021, pp. 1833–1844

    work page 2021

  31. [31]

    Z. Wang, X. Cun, J. Bao, W. Zhou, J. Liu, H. Li, Uformer: A general u-shaped transformer for image restoration, in: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2022, pp. 17683–17693

    work page 2022

  32. [32]

    S. W. Zamir, A. Arora, S. Khan, M. Hayat, F. S. Khan, M.-H. Yang, Restormer: Efficient transformer for high-resolution image restoration, in: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2022, pp. 5728–5739

    work page 2022

  33. [33]

    S. W. Zamir, A. Arora, S. Khan, M. Hayat, F. S. Khan, M.-H. Yang, L. Shao, Multi- stage progressive image restoration, in: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2021, pp. 14821–14831

    work page 2021

  34. [34]

    Kupyn, T

    O. Kupyn, T. Martyniuk, J. Wu, Z. Wang, Deblurgan-v2: Deblurring (orders- of-magnitude) faster and better, in: Proceedings of the IEEE/CVF international conference on computer vision, 2019, pp. 8878–8887

    work page 2019

  35. [35]

    B. Xia, Y . Zhang, S. Wang, Y . Wang, X. Wu, Y . Tian, W. Yang, L. Van Gool, Diffir: Efficient diffusion model for image restoration, in: Proceedings of the IEEE/CVF International Conference on Computer Vision, 2023, pp. 13095–13105. 30

    work page 2023

  36. [36]

    Z. Lu, Y . Liu, M. Jin, X. Luo, H. Yue, Z. Wang, S. Zuo, Y . Zeng, J. Fan, Y . Pang, et al., Virtual-scanning light-field microscopy for robust snapshot high-resolution volumetric imaging, Nature Methods 20 (5) (2023) 735–746

    work page 2023

  37. [37]

    X. Qin, Z. Wang, Y . Bai, X. Xie, H. Jia, Ffa-net: Feature fusion attention network for single image dehazing, in: Proceedings of the AAAI conference on artificial intelligence, V ol. 34, 2020, pp. 11908–11915

    work page 2020

  38. [38]

    Y . Song, Z. He, H. Qian, X. Du, Vision transformers for single image dehazing, IEEE Transactions on Image Processing 32 (2023) 1927–1941

    work page 2023

  39. [39]

    Yinglong, H

    W. Yinglong, H. Bin, Casdyf-net: Image dehazing via cascaded dynamic filters, arXiv preprint arXiv:2409.08510 (2024)

    work page arXiv 2024

  40. [40]

    B. Lim, S. Son, H. Kim, S. Nah, K. Mu Lee, Enhanced deep residual networks for single image super-resolution, in: Proceedings of the IEEE conference on computer vision and pattern recognition workshops, 2017, pp. 136–144

    work page 2017

  41. [41]

    X. Wang, L. Xie, C. Dong, Y . Shan, Real-esrgan: Training real-world blind super- resolution with pure synthetic data, in: Proceedings of the IEEE/CVF international conference on computer vision, 2021, pp. 1905–1914

    work page 2021

  42. [42]

    Hsu, C.-M

    C.-C. Hsu, C.-M. Lee, Y .-S. Chou, Drct: Saving image super-resolution away from information bottleneck, arXiv preprint arXiv:2404.00722 (2024)

    work page arXiv 2024

  43. [43]

    J. Su, B. Xu, H. Yin, A survey of deep learning approaches to image restoration, Neurocomputing 487 (2022) 46–65

    work page 2022

  44. [44]

    X. Mao, C. Shen, Y .-B. Yang, Image restoration using very deep convolutional encoder-decoder networks with symmetric skip connections, Advances in neural information processing systems 29 (2016)

    work page 2016

  45. [45]

    Zhang, K

    Y . Zhang, K. Li, K. Li, L. Wang, B. Zhong, Y . Fu, Image super-resolution using very deep residual channel attention networks, in: Proceedings of the European conference on computer vision (ECCV), 2018, pp. 286–301. 31

    work page 2018

  46. [46]

    Wang, J.-S

    D. Wang, J.-S. Pan, J.-H. Tang, Single image deraining using residual channel attention networks, Journal of Computer Science and Technology 38 (2) (2023) 439–454

    work page 2023

  47. [47]

    S. Guo, H. Yong, X. Zhang, J. Ma, L. Zhang, Spatial-frequency attention for image denoising, arXiv preprint arXiv:2302.13598 (2023)

    work page arXiv 2023

  48. [48]

    Purohit, M

    K. Purohit, M. Suin, A. Rajagopalan, V . N. Boddeti, Spatially-adaptive image restoration using distortion-guided networks, in: Proceedings of the IEEE/CVF international conference on computer vision, 2021, pp. 2309–2319

    work page 2021

  49. [49]

    Zhang, D

    Y . Zhang, D. Li, X. Shi, D. He, K. Song, X. Wang, H. Qin, H. Li, Kbnet: Kernel basis network for image restoration, arXiv preprint arXiv:2303.02881 (2023)

    work page arXiv 2023

  50. [50]

    Residual Non-local Attention Networks for Image Restoration

    Y . Zhang, K. Li, K. Li, B. Zhong, Y . Fu, Residual non-local attention networks for image restoration, arXiv preprint arXiv:1903.10082 (2019)

    work page internal anchor Pith review Pith/arXiv arXiv 1903

  51. [51]

    S. W. Zamir, A. Arora, S. Khan, M. Hayat, F. S. Khan, M.-H. Yang, L. Shao, Learn- ing enriched features for real image restoration and enhancement, in: Computer Vision–ECCV 2020: 16th European Conference, Glasgow, UK, August 23–28, 2020, Proceedings, Part XXV 16, Springer, 2020, pp. 492–511

    work page 2020

  52. [52]

    J. Liu, Q. Wang, H. Fan, Y . Wang, Y . Tang, L. Qu, Residual denoising diffusion models, in: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2024, pp. 2773–2783

    work page 2024

  53. [53]

    Z. Yue, K. Liao, C. C. Loy, Arbitrary-steps image super-resolution via diffusion inversion, arXiv preprint arXiv:2412.09013 (2024)

    work page arXiv 2024

  54. [54]

    J. W. Goodman, Introduction to Fourier Optics, Roberts and Company Publishers, 2005

    work page 2005

  55. [55]

    R. J. Noll, Zernike polynomials and atmospheric turbulence, JOsA 66 (3) (1976) 207–211. 32

    work page 1976

  56. [56]

    D. Saha, U. Schmidt, Q. Zhang, A. Barbotin, Q. Hu, N. Ji, M. J. Booth, M. Weigert, E. W. Myers, Practical sensorless aberration estimation for 3d microscopy with deep learning, Optics express 28 (20) (2020) 29044–29053

    work page 2020

  57. [57]

    Y . E. Kok, A. Bentley, A. J. Parkes, M. G. Somekh, A. J. Wright, M. P. Pound, Prac- tical aberration correction using deep transfer learning with limited experimental data, Optics Express 33 (6) (2025) 14431–14444

    work page 2025

  58. [58]

    Velickovic, G

    P. Velickovic, G. Cucurull, A. Casanova, A. Romero, P. Lio, Y . Bengio, et al., Graph attention networks, stat 1050 (20) (2017) 10–48550

    work page 2017

  59. [59]

    B. Deng, S. Song, A. P. French, D. Schluppeck, M. P. Pound, Advancing saliency ranking with human fixations: Dataset models and benchmarks, in: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2024, pp. 28348–28357

    work page 2024

  60. [60]

    Williams, J

    E. Williams, J. Moore, S. W. Li, G. Rustici, A. Tarkowska, A. Chessel, S. Leo, B. Antal, R. K. Ferguson, U. Sarkans, et al., Image data resource: a bioimage data integration and publication platform, Nature methods 14 (8) (2017) 775–781

    work page 2017

  61. [61]

    Ljosa, K

    V . Ljosa, K. L. Sokolnicki, A. E. Carpenter, Annotated high-throughput microscopy image sets for validation., Nature methods 9 (7) (2012) 637–637

    work page 2012

  62. [62]

    Huynh-Thu, M

    Q. Huynh-Thu, M. Ghanbari, Scope of validity of psnr in image/video quality assessment, Electronics letters 44 (13) (2008) 800–801

    work page 2008

  63. [63]

    Z. Wang, A. C. Bovik, H. R. Sheikh, E. P. Simoncelli, Image quality assessment: from error visibility to structural similarity, IEEE transactions on image processing 13 (4) (2004) 600–612

    work page 2004

  64. [64]

    Zhang, P

    R. Zhang, P. Isola, A. A. Efros, E. Shechtman, 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

    work page 2018

  65. [65]

    URLhttps://github.com/MrYxJ/calculate-flops.pytorch 33

    xiaoju ye, calflops: a flops and params calculate tool for neural networks in pytorch framework (2023). URLhttps://github.com/MrYxJ/calculate-flops.pytorch 33

    work page 2023

  66. [66]

    Maréchal, Study of the combined effects of diffraction and geometrical aber- rations on the image of a luminous point, Rev

    A. Maréchal, Study of the combined effects of diffraction and geometrical aber- rations on the image of a luminous point, Rev. Opt. Theor. Instrum 26 (1947) 257

    work page 1947

  67. [67]

    Mehri, P

    A. Mehri, P. B. Ardakani, A. D. Sappa, Mprnet: Multi-path residual network for lightweight image super resolution, in: Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision, 2021, pp. 2704–2713

    work page 2021

  68. [68]

    Ghasemabadi, M

    A. Ghasemabadi, M. K. Janjua, M. Salameh, C. Zhou, F. Sun, D. Niu, Cascad- edgaze: Efficiency in global context extraction for image restoration, arXiv preprint arXiv:2401.15235 (2024)

    work page arXiv 2024

  69. [69]

    Z. Liu, Y . Lin, Y . Cao, H. Hu, Y . Wei, Z. Zhang, S. Lin, B. Guo, Swin transformer: Hierarchical vision transformer using shifted windows, in: Proceedings of the IEEE/CVF international conference on computer vision, 2021, pp. 10012–10022. 34

    work page 2021