pith. sign in

arxiv: 2605.24959 · v1 · pith:FDDG5M3Fnew · submitted 2026-05-24 · 💻 cs.CV

Three-Step Conditional Diffusion 3D Reconstruction for Light-Field Microscopy

Qihong Zhao , Shaokang Yan , Zhimin Qiao , Jinjia Wang , Bo Xiong This is my paper

Pith reviewed 2026-06-30 11:40 UTC · model grok-4.3

classification 💻 cs.CV
keywords light-field microscopy3D reconstructionconditional diffusionvolumetric imagingdiffusion samplingbiological imaginganomaly detection
0
0 comments X

The pith

A three-step conditional diffusion model reconstructs 3D volumes from light-field microscopy with higher fidelity and better generalization than prior methods.

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

The paper proposes redesigning diffusion models for light-field microscopy 3D reconstruction by replacing the usual slow iterative sampling with a deterministic three-step process inside a lightweight conditional U-Net. This change is meant to remove the typical trade-off between reconstruction quality and computational speed while adding an Inter-Class Detection module to flag out-of-distribution inputs. The authors show through experiments that the resulting TCD method exceeds existing techniques in accuracy and in how well it works on data from different sources. If correct, the approach would make high-quality single-shot volumetric imaging practical for real-time biological observation.

Core claim

The authors establish that their Three-Step Conditional Diffusion (TCD) method, built on a deterministic three-step sampling strategy paired with a lightweight conditional U-Net and an Inter-Class Detection module, produces higher-fidelity 3D reconstructions and stronger cross-dataset generalization than state-of-the-art alternatives for light-field microscopy.

What carries the argument

The Three-Step Conditional Diffusion (TCD) process, which applies deterministic three-step sampling to a conditional U-Net together with the Inter-Class Detection (ICD) module that identifies out-of-distribution inputs.

If this is right

  • The three-step sampling removes the quality-efficiency trade-off that normally limits diffusion models in real-time imaging.
  • Cross-dataset results indicate the model maintains accuracy when input distributions shift.
  • The ICD module adds reliability by detecting anomalous samples at inference time.
  • Overall the method supplies a practical route to efficient high-fidelity volumetric reconstruction from single light-field captures.

Where Pith is reading between the lines

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

  • If the deterministic reduction works here, similar fixed-step sampling may be testable on conditioned diffusion tasks in other imaging domains.
  • Success would imply that full iterative diffusion is often unnecessary once strong conditioning is available.
  • The approach could support closed-loop live imaging setups where 3D volumes must be produced and acted on within seconds.

Load-bearing premise

A fixed deterministic three-step sampling schedule inside a lightweight conditional network can deliver the same or better reconstruction quality as full diffusion while working reliably on varied biological samples.

What would settle it

A controlled test on a new dataset in which TCD shows measurably lower reconstruction accuracy, slower inference, or poorer generalization than current leading methods would falsify the performance claims.

Figures

Figures reproduced from arXiv: 2605.24959 by Bo Xiong, Jinjia Wang, Qihong Zhao, Shaokang Yan, Zhimin Qiao.

Figure 1
Figure 1. Figure 1: Comparison of our TCD model with representa￾tive imaging methods: (a) Physics-based algorithms: Based on wave optics, these methods reconstruct 3D volumes by utilizing the system’s point spread function (PSF) matrix. (b) Learning￾based methods: Deep neural networks employ supervised learn￾ing to map light-field images to 3D volumes. (c) DDPM-based diffusion model: This method trains a U-Net-based one-step … view at source ↗
Figure 2
Figure 2. Figure 2: Overview of the U-Net-based one-step noise estimation framework. (a) The main U-Net architecture, composed of a sequence of Downsampling Blocks (DBs), Upsampling Blocks (UBs), and a Middle Block (MB) that includes a self-attention (Atten) module. Injectors are integrated within these blocks to enhance reconstruction fidelity. To reduce model complexity, select blocks are pruned. (b) The Injector module. Th… view at source ↗
Figure 3
Figure 3. Figure 3: A schematic overview of the complete workflow incorporating ICD detection. The figure illustrates how out-of-distribution samples are identified and subsequently handled. While conventional DDPM frameworks train only the noise estimator \protect \boldsymbol {\epsilon }_{\mathbf {\theta }} , our model optimizes the entire sampling trajectory by jointly minimizing the reconstruction loss be￾tween the final p… view at source ↗
Figure 4
Figure 4. Figure 4: Qualitative comparisons on the Vessel, Dendrite, and Mito datasets. where µ and Σ denote the mean vector and covariance ma￾trix of inter-class features, respectively. These parameters are fixed after training and serve as a baseline for anomaly detection during inference. During testing, yt−1 is extracted for each input sample, and its Mahalanobis distance to the in-distribution (ID) is computed [41]: \tex… view at source ↗
Figure 6
Figure 6. Figure 6: Framework comparison between TCD+ICD (left) and LFMNet+ICD (right). as ID samples. Experimental results demonstrate that TCD+ICD effec￾tively distinguishes most ID and OOD samples, whereas LFMNet+ICD fails to achieve stable recognition perfor￾mance, showing difficulty in separating these categories. This finding further verifies the high compatibility and synergy between the ICD module and the proposed TCD… view at source ↗
read the original abstract

Light-field microscopy (LFM) enables single-shot capture of multi-angular information from biological samples, supporting real-time volumetric imaging. However, traditional physics-based algorithms often suffer from limited spatial resolution, severe artifacts, and high computational costs. Existing learning-based methods improve inference efficiency but still face limitations in reconstruction accuracy and generalization capability. To address these challenges, this paper proposes a high-fidelity Three-Step Conditional Diffusion (TCD) 3D reconstruction method for LFM. Although conventional diffusion models have achieved remarkable success in generative modeling, their slow sampling process and the inherent trade-off between quality and efficiency hinder their application in real-time 3D imaging. We redesign the diffusion process through a deterministic three-step sampling strategy coupled with a lightweight conditional U-Net, establishing a new paradigm for fast and accurate volumetric reconstruction. Furthermore, an Inter-Class Detection (ICD) module is incorporated to identify out-of-distribution or anomalous inputs during inference, thereby enhancing model stability and reliability. Extensive experiments and cross-dataset evaluations demonstrate that TCD significantly outperforms state-of-the-art methods in both reconstruction fidelity and generalization, providing an efficient and practical 3D reconstruction solution for light-field microscopy.

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

1 major / 0 minor

Summary. The paper proposes a Three-Step Conditional Diffusion (TCD) method for 3D reconstruction in light-field microscopy. It introduces a deterministic three-step sampling strategy paired with a lightweight conditional U-Net and an Inter-Class Detection (ICD) module to address the quality-efficiency trade-off in diffusion models, claiming improved reconstruction fidelity, efficiency, and generalization across datasets via extensive experiments and cross-dataset evaluations.

Significance. If the experimental claims hold, the work could offer a practical advance for real-time volumetric imaging in biological applications by providing an efficient alternative to both physics-based and existing learning-based LFM reconstruction methods. No mention of machine-checked proofs, reproducible code releases, or parameter-free derivations is present in the provided text.

major comments (1)
  1. Abstract: the central claim that TCD 'significantly outperforms state-of-the-art methods in both reconstruction fidelity and generalization' is asserted without any quantitative results, error metrics, baselines, ablation studies, or method implementation details supplied in the available text, rendering the claim unverifiable from the manuscript as presented.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the detailed review and the opportunity to clarify our presentation. We address the major comment below.

read point-by-point responses

  1. Referee: Abstract: the central claim that TCD 'significantly outperforms state-of-the-art methods in both reconstruction fidelity and generalization' is asserted without any quantitative results, error metrics, baselines, ablation studies, or method implementation details supplied in the available text, rendering the claim unverifiable from the manuscript as presented.

    Authors: The full manuscript includes a dedicated Experiments section (Section 4) that reports quantitative results using standard metrics such as PSNR, SSIM, and MAE, with direct comparisons against multiple state-of-the-art baselines (both physics-based and learning-based), ablation studies on the three-step sampling, conditional U-Net, and ICD module, cross-dataset generalization tests, and full implementation details including network architecture and training protocols. The abstract is intended as a concise summary of these findings. If the version provided to the referee contained only the abstract, we can supply the complete manuscript or, if preferred, revise the abstract to include one or two key quantitative highlights while respecting length constraints. revision: partial

Circularity Check

0 steps flagged

No significant circularity detected

full rationale

The abstract and available description present a methodological proposal (deterministic three-step sampling with conditional U-Net and ICD module) whose central claims rest on asserted experimental outcomes rather than any closed derivation chain. No equations, parameter fits, self-citations, or uniqueness theorems are exhibited that would allow a reduction of outputs to inputs by construction. The reader's assessment of zero circularity is therefore confirmed; the work is treated as self-contained against external benchmarks until full equations are inspected.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract supplies no information on free parameters, axioms, or invented entities; full text required for audit.

pith-pipeline@v0.9.1-grok · 5742 in / 981 out tokens · 39902 ms · 2026-06-30T11:40:04.451965+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

47 extracted references · 3 canonical work pages · 2 internal anchors

  1. [1]

    Fast near-whole–brain imaging in adult drosophila during responses to stimuli and behavior.PLoS biology, 17(2):e2006732, 2019

    Sophie Aimon, Takeo Katsuki, Tongqiu Jia, Logan Grosenick, Michael Broxton, Karl Deisseroth, Terrence J Sejnowski, and Ralph J Greenspan. Fast near-whole–brain imaging in adult drosophila during responses to stimuli and behavior.PLoS biology, 17(2):e2006732, 2019. 2

    2019

  2. [2]

    The light field camera: Extended depth of field, aliasing, and superresolution.IEEE transactions on pattern analysis and machine intelligence, 34(5):972–986, 2011

    Tom E Bishop and Paolo Favaro. The light field camera: Extended depth of field, aliasing, and superresolution.IEEE transactions on pattern analysis and machine intelligence, 34(5):972–986, 2011. 2

    2011

  3. [3]

    Wave optics theory and 3-d deconvolution for the light field microscope.Optics express, 21(21):25418–25439, 2013

    Michael Broxton, Logan Grosenick, Samuel Yang, Noy Co- hen, Aaron Andalman, Karl Deisseroth, and Marc Levoy. Wave optics theory and 3-d deconvolution for the light field microscope.Optics express, 21(21):25418–25439, 2013. 1, 2, 7

    2013

  4. [4]

    Invertible diffusion models for compressed sensing.IEEE Transactions on Pat- tern Analysis and Machine Intelligence, 2025

    Bin Chen, Zhenyu Zhang, Weiqi Li, Chen Zhao, Jiwen Yu, Shijie Zhao, Jie Chen, and Jian Zhang. Invertible diffusion models for compressed sensing.IEEE Transactions on Pat- tern Analysis and Machine Intelligence, 2025. 2

    2025

  5. [5]

    Three-dimensional residual channel attention networks denoise and sharpen flu- orescence microscopy image volumes.Nature methods, 18 (6):678–687, 2021

    Jiji Chen, Hideki Sasaki, Hoyin Lai, Yijun Su, Jiamin Liu, Yicong Wu, Alexander Zhovmer, Christian A Combs, Ivan Rey-Suarez, Hung-Yu Chang, et al. Three-dimensional residual channel attention networks denoise and sharpen flu- orescence microscopy image volumes.Nature methods, 18 (6):678–687, 2021. 2

    2021

  6. [6]

    Opportunities and challenges of diffusion models for gener- ative ai.National Science Review, 11(12):nwae348, 2024

    Minshuo Chen, Song Mei, Jianqing Fan, and Mengdi Wang. Opportunities and challenges of diffusion models for gener- ative ai.National Science Review, 11(12):nwae348, 2024. 3

    2024

  7. [7]

    Optical-aberrations- corrected light field re-projection for high-quality plenoptic imaging.Optics Express, 28(3):3057–3072, 2020

    Yanqin Chen, Xin Jin, and Bo Xiong. Optical-aberrations- corrected light field re-projection for high-quality plenoptic imaging.Optics Express, 28(3):3057–3072, 2020. 2

    2020

  8. [8]

    Rapid whole brain imaging of neural ac- tivity in freely behaving larval zebrafish (danio rerio).elife, 6:e28158, 2017

    Lin Cong, Zeguan Wang, Yuming Chai, Wei Hang, Chun- feng Shang, Wenbin Yang, Lu Bai, Jiulin Du, Kai Wang, and Quan Wen. Rapid whole brain imaging of neural ac- tivity in freely behaving larval zebrafish (danio rerio).elife, 6:e28158, 2017. 1

    2017

  9. [9]

    Diffusion models beat gans on image synthesis.Advances in neural informa- tion processing systems, 34:8780–8794, 2021

    Prafulla Dhariwal and Alexander Nichol. Diffusion models beat gans on image synthesis.Advances in neural informa- tion processing systems, 34:8780–8794, 2021. 3

    2021

  10. [10]

    Don’t drop your samples! coherence-aware training benefits conditional diffusion

    Nicolas Dufour, Victor Besnier, Vicky Kalogeiton, and David Picard. Don’t drop your samples! coherence-aware training benefits conditional diffusion. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 6264–6273, 2024. 3

    2024

  11. [11]

    Out-of-distribution (ood) detection and generalization improved by augmenting adversarial mixup samples.Electronics, 12(6):1421, 2023

    Kyungpil Gwon and Joonhyuk Yoo. Out-of-distribution (ood) detection and generalization improved by augmenting adversarial mixup samples.Electronics, 12(6):1421, 2023. 2

    2023

  12. [12]

    Denoising dif- fusion probabilistic models.Advances in neural information processing systems, 33:6840–6851, 2020

    Jonathan Ho, Ajay Jain, and Pieter Abbeel. Denoising dif- fusion probabilistic models.Advances in neural information processing systems, 33:6840–6851, 2020. 2, 3, 7

    2020

  13. [13]

    A simple unified framework for detecting out-of-distribution samples and adversarial attacks.Advances in neural infor- mation processing systems, 31, 2018

    Kimin Lee, Kibok Lee, Honglak Lee, and Jinwoo Shin. A simple unified framework for detecting out-of-distribution samples and adversarial attacks.Advances in neural infor- mation processing systems, 31, 2018. 3

    2018

  14. [14]

    Light field microscopy

    Marc Levoy, Ren Ng, Andrew Adams, Matthew Footer, and Mark Horowitz. Light field microscopy. InAcm siggraph 2006 papers, pages 924–934. 2006. 2

    2006

  15. [15]

    Record- ing and controlling the 4d light field in a microscope using microlens arrays.Journal of microscopy, 235(2):144–162,

    Marc Levoy, Zhengyun Zhang, and Ian McDowall. Record- ing and controlling the 4d light field in a microscope using microlens arrays.Journal of microscopy, 235(2):144–162,

  16. [16]

    Faster diffusion: Rethinking the role of the encoder for diffusion model inference.Advances in Neural Information Processing Systems, 37:85203–85240,

    Senmao Li, Taihang Hu, Joost van de Weijer, Fahad S Khan, Tao Liu, Linxuan Li, Shiqi Yang, Yaxing Wang, Ming-Ming Cheng, and Jian Yang. Faster diffusion: Rethinking the role of the encoder for diffusion model inference.Advances in Neural Information Processing Systems, 37:85203–85240,

  17. [17]

    Incorporating the image formation process into deep learning improves network per- formance.Nature Methods, 19(11):1427–1437, 2022

    Yue Li, Yijun Su, Min Guo, Xiaofei Han, Jiamin Liu, Har- shad D Vishwasrao, Xuesong Li, Ryan Christensen, Titas Sengupta, Mark W Moyle, et al. Incorporating the image formation process into deep learning improves network per- formance.Nature Methods, 19(11):1427–1437, 2022. 2

    2022

  18. [18]

    Enhancing the reliability of out-of-distribution image detection in neural networks.arXiv preprint arXiv:1706.02690,

    Shiyu Liang, Yixuan Li, and Rayadurgam Srikant. Enhanc- ing the reliability of out-of-distribution image detection in neural networks.arXiv preprint arXiv:1706.02690, 2017. 3

    work page arXiv 2017

  19. [19]

    Residual denoising diffu- sion models

    Jiawei Liu, Qiang Wang, Huijie Fan, Yinong Wang, Yan- dong Tang, and Liangqiong Qu. Residual denoising diffu- sion models. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 2773– 2783, 2024. 2, 3

    2024

  20. [20]

    Phase- space deconvolution for light field microscopy.Optics ex- press, 27(13):18131–18145, 2019

    Zhi Lu, Jiamin Wu, Hui Qiao, You Zhou, Tao Yan, Zijing Zhou, Xu Zhang, Jingtao Fan, and Qionghai Dai. Phase- space deconvolution for light field microscopy.Optics ex- press, 27(13):18131–18145, 2019. 2

    2019

  21. [21]

    Conditional image-to-video gener- ation with latent flow diffusion models

    Haomiao Ni, Changhao Shi, Kai Li, Sharon X Huang, and Martin Renqiang Min. Conditional image-to-video gener- ation with latent flow diffusion models. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 18444–18455, 2023. 3

    2023

  22. [22]

    Fast light-field 3d microscopy with out-of- distribution detection and adaptation through conditional normalizing flows.Biomedical Optics Express, 15(2):1219– 1232, 2024

    Josu ´e Page Vizca´ıno, Panagiotis Symvoulidis, Zeguan Wang, Jonas Jelten, Paolo Favaro, Edward S Boyden, and To- bias Lasser. Fast light-field 3d microscopy with out-of- distribution detection and adaptation through conditional normalizing flows.Biomedical Optics Express, 15(2):1219– 1232, 2024. 2, 7

    2024

  23. [23]

    Simultaneous whole-animal 3d imaging of neuronal activity using light-field microscopy.Nature methods, 11(7):727– 730, 2014

    Robert Prevedel, Young-Gyu Yoon, Maximilian Hoffmann, Nikita Pak, Gordon Wetzstein, Saul Kato, Tina Schr ¨odel, Ramesh Raskar, Manuel Zimmer, Edward S Boyden, et al. Simultaneous whole-animal 3d imaging of neuronal activity using light-field microscopy.Nature methods, 11(7):727– 730, 2014. 2

    2014

  24. [24]

    Deep-learning-based noise correction method for light-field fluorescence microscopy.Photonics Research, 13(9):2547– 2565, 2025

    Bohan Qu, Zhouyu Jin, You Zhou, Bo Xiong, and Xun Cao. Deep-learning-based noise correction method for light-field fluorescence microscopy.Photonics Research, 13(9):2547– 2565, 2025. 1

    2025

  25. [25]

    High-resolution image synthesis with latent diffusion models

    Robin Rombach, Andreas Blattmann, Dominik Lorenz, Patrick Esser, and Bj ¨orn Ommer. High-resolution image synthesis with latent diffusion models. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 10684–10695, 2022. 3

    2022

  26. [26]

    Mm-diffusion: Learning multi-modal diffusion mod- els for joint audio and video generation

    Ludan Ruan, Yiyang Ma, Huan Yang, Huiguo He, Bei Liu, Jianlong Fu, Nicholas Jing Yuan, Qin Jin, and Baining Guo. Mm-diffusion: Learning multi-modal diffusion mod- els for joint audio and video generation. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 10219–10228, 2023. 3

    2023

  27. [27]

    Photorealistic text-to-image diffusion models with deep language understanding.Advances in neural information processing systems, 35:36479–36494, 2022

    Chitwan Saharia, William Chan, Saurabh Saxena, Lala Li, Jay Whang, Emily L Denton, Kamyar Ghasemipour, Raphael Gontijo Lopes, Burcu Karagol Ayan, Tim Salimans, et al. Photorealistic text-to-image diffusion models with deep language understanding.Advances in neural information processing systems, 35:36479–36494, 2022. 3

    2022

  28. [28]

    Advances in two-photon scanning and scanless microscopy technologies for func- tional neural circuit imaging.Proceedings of the IEEE, 105 (1):139–157, 2016

    Simon R Schultz, Caroline S Copeland, Amanda J Foust, Pe- ter Quicke, and Renaud Schuck. Advances in two-photon scanning and scanless microscopy technologies for func- tional neural circuit imaging.Proceedings of the IEEE, 105 (1):139–157, 2016. 2

    2016

  29. [29]

    Image formation anal- ysis and high resolution image reconstruction for plenoptic imaging systems.Applied optics, 52(10):D22–D31, 2013

    Sapna A Shroff and Kathrin Berkner. Image formation anal- ysis and high resolution image reconstruction for plenoptic imaging systems.Applied optics, 52(10):D22–D31, 2013. 1, 2

    2013

  30. [30]

    Deep unsupervised learning using nonequilibrium thermodynamics

    Jascha Sohl-Dickstein, Eric Weiss, Niru Maheswaranathan, and Surya Ganguli. Deep unsupervised learning using nonequilibrium thermodynamics. InInternational confer- ence on machine learning, pages 2256–2265. pmlr, 2015. 3

    2015

  31. [31]

    Score-Based Generative Modeling through Stochastic Differential Equations

    Yang Song, Jascha Sohl-Dickstein, Diederik P Kingma, Ab- hishek Kumar, Stefano Ermon, and Ben Poole. Score-based generative modeling through stochastic differential equa- tions.arXiv preprint arXiv:2011.13456, 2020. 3

    work page internal anchor Pith review Pith/arXiv arXiv 2011

  32. [32]

    Artifact-free deconvolution in light field microscopy.Optics express, 27(22):31644– 31666, 2019

    Anca Stefanoiu, Josue Page, Panagiotis Symvoulidis, Gil G Westmeyer, and Tobias Lasser. Artifact-free deconvolution in light field microscopy.Optics express, 27(22):31644– 31666, 2019. 2

    2019

  33. [33]

    3d deconvolu- tion in fourier integral microscopy

    Anca Stefanoiu, Gabriele Scrofani, Genaro Saavedra, Manuel Mart´ınez-Corral, and Tobias Lasser. 3d deconvolu- tion in fourier integral microscopy. InComputational Imag- ing V, pages 56–63. SPIE, 2020. 2

    2020

  34. [34]

    Autodeconj: a gpu- accelerated imagej plugin for 3d light-field deconvolution with optimal iteration numbers predicting.Bioinformatics, 39(1):btac760, 2023

    Changqing Su, Yuhan Gao, You Zhou, Yaoqi Sun, Cheng- gang Yan, Haibing Yin, and Bo Xiong. Autodeconj: a gpu- accelerated imagej plugin for 3d light-field deconvolution with optimal iteration numbers predicting.Bioinformatics, 39(1):btac760, 2023. 2

    2023

  35. [35]

    Generaliza- tion of neural network models for complex network dynam- ics.Communications Physics, 7(1):348, 2024

    Vaiva Vasiliauskaite and Nino Antulov-Fantulin. Generaliza- tion of neural network models for complex network dynam- ics.Communications Physics, 7(1):348, 2024. 3

    2024

  36. [36]

    V olume recon- struction for light field microscopy

    Herman Verinaz-Jadan, Pingfan Song, Carmel L Howe, Amanda J Foust, and Pier Luigi Dragotti. V olume recon- struction for light field microscopy. InICASSP 2020-2020 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), pages 1459–1463. IEEE, 2020. 1, 2

    2020

  37. [37]

    Shift-invariant- subspace discretization and volume reconstruction for light field microscopy.IEEE Transactions on Computational Imaging, 8:286–301, 2022

    Herman Verinaz-Jadan, Pingfan Song, Carmel L Howe, Amanda J Foust, and Pier Luigi Dragotti. Shift-invariant- subspace discretization and volume reconstruction for light field microscopy.IEEE Transactions on Computational Imaging, 8:286–301, 2022. 1, 2

    2022

  38. [38]

    Learning to recon- struct confocal microscopy stacks from single light field im- ages.IEEE transactions on computational imaging, 7:775– 788, 2021

    Josue Page Vizcaino, Federico Saltarin, Yury Belyaev, Ruth Lyck, Tobias Lasser, and Paolo Favaro. Learning to recon- struct confocal microscopy stacks from single light field im- ages.IEEE transactions on computational imaging, 7:775– 788, 2021. 2, 7

    2021

  39. [39]

    Real-time volumetric reconstruction of biological dynamics with light-field microscopy and deep learning.Nature methods, 18(5):551–556, 2021

    Zhaoqiang Wang, Lanxin Zhu, Hao Zhang, Guo Li, Chengqiang Yi, Yi Li, Yicong Yang, Yichen Ding, Mei Zhen, Shangbang Gao, et al. Real-time volumetric reconstruction of biological dynamics with light-field microscopy and deep learning.Nature methods, 18(5):551–556, 2021. 2, 7

    2021

  40. [40]

    3D Gaussian Adaptive Reconstruction for Fourier Light-Field Microscopy

    Chenyu Xu, Zhouyu Jin, Chengkang Shen, Hao Zhu, Zhan Ma, Bo Xiong, You Zhou, Xun Cao, and Ning Gu. 3d gaussian adaptive reconstruction for fourier light-field mi- croscopy.arXiv preprint arXiv:2505.12875, 2025. 1

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  41. [41]

    Out-of-distribution detection based on distance metric learning

    Donghun Yang, Iksoo Shin, Mai Ngoc Kien, Hoyong Kim, Chanhee Yu, and Myunggwon Hwang. Out-of-distribution detection based on distance metric learning. InThe 9th In- ternational Conference on Smart Media and Applications, pages 214–218, 2020. 3, 6

    2020

  42. [42]

    Light field microscopy in biological imaging.Journal of Innova- tive Optical Health Sciences, 16(01):2230017, 2023

    Chengqiang Yi, Lanxin Zhu, Dongyu Li, and Peng Fei. Light field microscopy in biological imaging.Journal of Innova- tive Optical Health Sciences, 16(01):2230017, 2023. 1

    2023

  43. [43]

    Adding conditional control to text-to-image diffusion models

    Lvmin Zhang, Anyi Rao, and Maneesh Agrawala. Adding conditional control to text-to-image diffusion models. In Proceedings of the IEEE/CVF international conference on computer vision, pages 3836–3847, 2023. 3

    2023

  44. [44]

    V2v3d: View-to-view denoised 3d reconstruction for light field mi- croscopy

    Jiayin Zhao, Zhenqi Fu, Tao Yu, and Hui Qiao. V2v3d: View-to-view denoised 3d reconstruction for light field mi- croscopy. InProceedings of the Computer Vision and Pattern Recognition Conference, pages 26451–26461, 2025. 2, 3

    2025

  45. [45]

    Sparse deconvolution im- proves the resolution of live-cell super-resolution fluores- cence microscopy.Nature biotechnology, 40(4):606–617,

    Weisong Zhao, Shiqun Zhao, Liuju Li, Xiaoshuai Huang, Shijia Xing, Yulin Zhang, Guohua Qiu, Zhenqian Han, Yingxu Shang, De-en Sun, et al. Sparse deconvolution im- proves the resolution of live-cell super-resolution fluores- cence microscopy.Nature biotechnology, 40(4):606–617,

  46. [46]

    Aberration modeling in deep learning for volumetric reconstruction of light-field microscopy.Laser & Photonics Reviews, 17(10):2300154, 2023

    You Zhou, Zhouyu Jin, Qianhui Zhao, Bo Xiong, and Xun Cao. Aberration modeling in deep learning for volumetric reconstruction of light-field microscopy.Laser & Photonics Reviews, 17(10):2300154, 2023. 1

    2023

  47. [47]

    Disorder- invariant implicit neural representation.IEEE Transactions on Pattern Analysis and Machine Intelligence, 46(8):5463– 5478, 2024

    Hao Zhu, Shaowen Xie, Zhen Liu, Fengyi Liu, Qi Zhang, You Zhou, Yi Lin, Zhan Ma, and Xun Cao. Disorder- invariant implicit neural representation.IEEE Transactions on Pattern Analysis and Machine Intelligence, 46(8):5463– 5478, 2024. 2

    2024