pith. sign in

arxiv: 2606.31959 · v1 · pith:VXR5TAXKnew · submitted 2026-06-30 · 💻 cs.CV

AnyBokeh: Physics-Guided Any-to-Any Bokeh Editing with Optical Fingerprint Transfer

Xinyu Hou , Xiaoming Li , Zongsheng Yue , Chen Change Loy This is my paper

Pith reviewed 2026-07-01 05:25 UTC · model grok-4.3

classification 💻 cs.CV
keywords bokeh editingcircle of confusionoptical fingerprintdefocus deblurringdepth of field controlgenerative editingphysics-guided synthesisany-to-any image editing
0
0 comments X

The pith

AnyBokeh performs any-to-any bokeh editing on single images by estimating and transferring source optical fingerprints through signed circle-of-confusion maps.

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

The paper introduces a method to edit bokeh in photographs captured under any focus and aperture conditions to any desired settings. Instead of first removing blur to create an all-in-focus image, it directly estimates the source's blur state using a signed circle-of-confusion map and a disparity map. This allows modeling how the source's optical characteristics relate to depth differences, creating a transferable fingerprint that guides editing to new focus and aperture values. A generative model then synthesizes the relative changes in blur while preserving useful cues from the original image. This approach avoids common steps that can introduce artifacts and enables more faithful results across editing, rendering, and deblurring tasks.

Core claim

AnyBokeh estimates a source-specific optical fingerprint by modeling the linear relation between signed circle-of-confusion and disparity difference from a single input image, then transfers this fingerprint using a generative editor conditioned on source and target circle-of-confusion maps to achieve relative blur synthesis for any-to-any bokeh editing.

What carries the argument

The source-specific optical fingerprint derived from the linear relation between signed circle-of-confusion and disparity difference, used to condition a generative editor for relative blur synthesis.

If this is right

  • Any-to-any bokeh editing becomes possible without requiring an all-in-focus input or reconstruction.
  • Defocus deblurring and new bokeh rendering preserve original optical characteristics.
  • No test-time bokeh-level calibration is needed for different source images.
  • A high-fidelity synthetic dataset with depth, focus distance, and EXIF metadata supports supervised training.

Where Pith is reading between the lines

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

  • Similar optical fingerprint transfer might apply to other lens effects such as chromatic aberration.
  • Real-time mobile photography apps could incorporate this for on-device bokeh adjustments.
  • The method suggests that preserving source blur cues improves consistency in edited images compared to reconstruction pipelines.

Load-bearing premise

A linear relation between signed circle-of-confusion and disparity difference exists and suffices to estimate a transferable source-specific optical fingerprint from a single image.

What would settle it

Real-world images where applying the transferred optical fingerprint produces inconsistent bokeh patterns that do not match ground-truth captures at the target focus and aperture settings.

Figures

Figures reproduced from arXiv: 2606.31959 by Chen Change Loy, Xiaoming Li, Xinyu Hou, Zongsheng Yue.

Figure 1
Figure 1. Figure 1: Overview of the AnyBokeh framework. (1) Stage 1 predicts the signed source CoC map [PITH_FULL_IMAGE:figures/full_fig_p005_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Samples from UnrealBokeh. For the same (environment, camera, focal length) scene, [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Qualitative Stage 1 predictions on EBB! and RealBokeh. Each row shows the same scene [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Qualitative comparison on any-to-any bokeh editing. Aperture values are shown for the [PITH_FULL_IMAGE:figures/full_fig_p009_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Qualitative results for arbitrary aperture and focus changes. Given a single input image, [PITH_FULL_IMAGE:figures/full_fig_p012_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Qualitative examples illustrating the gap between LPIPS [ [PITH_FULL_IMAGE:figures/full_fig_p014_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Visual comparisons for ablation studies. We evaluate the effect of the two-stage design and [PITH_FULL_IMAGE:figures/full_fig_p015_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Failure cases under extreme source blur. (1) When the input is severely defocused (first [PITH_FULL_IMAGE:figures/full_fig_p016_8.png] view at source ↗
read the original abstract

Depth-of-field control is a fundamental tool in photography, yet post-capture bokeh editing from a single image remains challenging. A practical editor should handle images captured under arbitrary focus and aperture settings. Existing methods typically assume an all-in-focus input, or first recover an all-in-focus image before rendering new bokeh. Such pipelines can discard useful blur cues from the source image and propagate reconstruction artifacts into the final edit. We introduce AnyBokeh, a physics-guided framework for any-to-any bokeh editing. Instead of treating source blur merely as a degradation to be removed, AnyBokeh estimates the source blur state with a signed circle-of-confusion map and a disparity map. By modeling the linear relation between signed circle of confusion and disparity difference, AnyBokeh estimates a source-specific optical fingerprint and transfers the source optical characteristics to the desired focus and aperture setting. A generative editor conditioned on both source and target circle-of-confusion maps then performs relative blur synthesis, enabling spatially adaptive deblurring, preservation, and defocus rendering. To support physically supervised learning, we further construct a high-fidelity synthetic dataset with accurate depth, focus distance, and full EXIF metadata. Experiments on real-world benchmarks show that AnyBokeh achieves faithful and controllable editing across any-to-any bokeh editing, all-in-focus-to-bokeh rendering, and defocus deblurring, while avoiding all-in-focus reconstruction and test-time bokeh-level calibration commonly required by existing approaches. The code and dataset will be available at https://github.com/itsmag11/AnyBokeh.

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 / 0 minor

Summary. The manuscript introduces AnyBokeh, a physics-guided framework for any-to-any bokeh editing from a single image. It estimates source blur state via a signed circle-of-confusion map and disparity map, models an assumed linear relation between signed circle-of-confusion and disparity difference to extract a source-specific optical fingerprint, transfers this fingerprint to target focus/aperture settings, and conditions a generative editor on source and target CoC maps for relative blur synthesis. The work also constructs a high-fidelity synthetic dataset with accurate depth, focus distance, and full EXIF metadata. Experiments on real-world benchmarks are claimed to demonstrate faithful and controllable editing across any-to-any bokeh editing, all-in-focus-to-bokeh rendering, and defocus deblurring, while avoiding all-in-focus reconstruction and test-time calibration.

Significance. If the linear signed-CoC/disparity relation holds with acceptable error across lens regimes and the generative editor generalizes, the approach could meaningfully advance practical single-image bokeh control by retaining rather than discarding source blur cues. The release of code and the synthetic dataset with full metadata is a clear strength for reproducibility.

major comments (2)
  1. [Abstract] Abstract: the central claim that a linear relation between signed circle-of-confusion and disparity difference enables reliable single-image extraction of a transferable optical fingerprint is stated without derivation, error bounds, or ablation on robustness to lens aberrations, depth errors, or non-paraxial effects. This assumption is load-bearing for the fingerprint-transfer pipeline and the claim of avoiding all-in-focus reconstruction.
  2. [Abstract] Abstract: performance claims of 'faithful and controllable editing' and 'superior performance' on real-world benchmarks are asserted without any quantitative metrics, error analysis, or comparison tables visible in the provided text, preventing assessment of the central empirical contribution.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback. Below we respond point-by-point to the major comments and indicate the revisions we will incorporate.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the central claim that a linear relation between signed circle-of-confusion and disparity difference enables reliable single-image extraction of a transferable optical fingerprint is stated without derivation, error bounds, or ablation on robustness to lens aberrations, depth errors, or non-paraxial effects. This assumption is load-bearing for the fingerprint-transfer pipeline and the claim of avoiding all-in-focus reconstruction.

    Authors: The abstract is necessarily concise. The derivation of the linear signed-CoC/disparity relation from the thin-lens equation appears in Section 3.2, together with the paraxial approximation used. We will revise the abstract to explicitly reference this derivation and the conditions under which the relation holds. In addition, we will add a dedicated ablation subsection (new Section 4.4) that quantifies error under depth noise, mild aberrations, and non-paraxial regimes, reporting the resulting fingerprint transfer error. These changes directly address the load-bearing nature of the assumption. revision: yes

  2. Referee: [Abstract] Abstract: performance claims of 'faithful and controllable editing' and 'superior performance' on real-world benchmarks are asserted without any quantitative metrics, error analysis, or comparison tables visible in the provided text, preventing assessment of the central empirical contribution.

    Authors: The abstract summarizes the experimental outcomes; the quantitative results (PSNR, SSIM, LPIPS, user-study scores, and comparison tables against prior any-to-any and defocus methods) are reported in Section 4 with error bars and statistical significance. To make the abstract self-contained, we will insert the key numerical improvements (e.g., “+1.8 dB PSNR over the strongest baseline on the RealBokeh set”) while remaining within length limits. This revision will allow readers to assess the empirical claims directly from the abstract. revision: yes

Circularity Check

0 steps flagged

No circularity; linear CoC-disparity relation is an explicit modeling assumption, not a self-derived result

full rationale

The paper's central step models a linear relation between signed circle-of-confusion and disparity difference to extract a source optical fingerprint from a single image. This is presented as a physics-based modeling choice (abstract: 'By modeling the linear relation... AnyBokeh estimates a source-specific optical fingerprint'), not derived from or equivalent to the target editing output by construction. No equations reduce the fingerprint or final edit to a fitted parameter defined from the input; the framework uses a new synthetic dataset for supervision and avoids all-in-focus reconstruction explicitly. No self-citation chains, uniqueness theorems, or renamed empirical patterns are load-bearing. The derivation chain remains independent of the output.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 2 invented entities

The central claim rests on the domain assumption of a linear signed-CoC-to-disparity relation and the introduction of the optical fingerprint concept; no free parameters or invented entities with independent evidence are detailed in the abstract.

axioms (1)
  • domain assumption Linear relation between signed circle of confusion and disparity difference
    Invoked to estimate source-specific optical fingerprint from the input image.
invented entities (2)
  • signed circle-of-confusion map no independent evidence
    purpose: Represent source blur state for fingerprint estimation
    New representation introduced to capture both magnitude and sign of defocus.
  • optical fingerprint no independent evidence
    purpose: Capture and transfer source-specific optical characteristics
    Core new construct enabling any-to-any transfer without reconstruction.

pith-pipeline@v0.9.1-grok · 5827 in / 1302 out tokens · 33762 ms · 2026-07-01T05:25:28.277539+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

44 extracted references

  1. [1]

    Defocus deblurring using dual-pixel data

    Abdullah Abuolaim and Michael S Brown. Defocus deblurring using dual-pixel data. InEuropean conference on computer vision (ECCV), 2020

    2020

  2. [2]

    Towards digital refocusing from a single photograph

    Yosuke Bando and Tomoyuki Nishita. Towards digital refocusing from a single photograph. InPacific Conference on Computer Graphics and Applications, pages 363–372, 2007

    2007

  3. [3]

    Barron, Andrew Adams, YiChang Shih, and Carlos Hernández

    Jonathan T. Barron, Andrew Adams, YiChang Shih, and Carlos Hernández. Fast bilateral-space stereo for synthetic defocus. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2015

    2015

  4. [4]

    Richter, and Vladlen Koltun

    Aleksei Bochkovskii, Amaël Delaunoy, Hugo Germain, Marcel Santos, Yichao Zhou, Stephan R. Richter, and Vladlen Koltun. Depth Pro: Sharp monocular metric depth in less than a second. InInternational Conference on Learning Representations (ICLR), 2025

    2025

  5. [5]

    SteReFo: Efficient image refocusing with stereo vision

    Benjamin Busam, Matthieu Hog, Steven McDonagh, and Gregory Slabaugh. SteReFo: Efficient image refocusing with stereo vision. InProceedings of the IEEE/CVF International Conference on Computer Vision Workshops (ICCVW), 2019

    2019

  6. [6]

    Quad- pixel image defocus deblurring: A new benchmark and model

    Hang Chen, Yin Xie, Xiaoxiu Peng, Lihu Sun, Wenkai Su, Xiaodong Yang, and Chengming Liu. Quad- pixel image defocus deblurring: A new benchmark and model. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2025

    2025

  7. [7]

    Simoncelli

    Keyan Ding, Kede Ma, Shiqi Wang, and Eero P. Simoncelli. Image quality assessment: Unifying structure and texture similarity.IEEE Transactions on Pattern Analysis and Machine Intelligence (PAMI), 44(5): 2567–2581, 2022

    2022

  8. [8]

    Unreal engine

    Epic Games. Unreal engine

  9. [9]

    Bokeh Diffusion: Defocus blur control in text-to-image diffusion models

    Armando Fortes, Tianyi Wei, Shangchen Zhou, and Xingang Pan. Bokeh Diffusion: Defocus blur control in text-to-image diffusion models. InProceedings of the SIGGRAPH Asia 2025 Conference Papers, 2025

    2025

  10. [10]

    Cinematic bokeh rendering for real scenes

    Thomas Hach, Johannes Steurer, Arvind Amruth, and Artur Pappenheim. Cinematic bokeh rendering for real scenes. InProceedings of the 12th European Conference on Visual Media Production, 2015

    2015

  11. [11]

    The accumulation buffer: hardware support for high-quality rendering

    Paul Haeberli and Kurt Akeley. The accumulation buffer: hardware support for high-quality rendering. International Conference on Computer Graphics and Interactive Techniques, 1990

    1990

  12. [12]

    Gans trained by a two time-scale update rule converge to a local nash equilibrium

    Martin Heusel, Hubert Ramsauer, Thomas Unterthiner, Bernhard Nessler, and Sepp Hochreiter. Gans trained by a two time-scale update rule converge to a local nash equilibrium. InAdvances in Neural Information Processing Systems, page 6629–6640, 2017

    2017

  13. [13]

    LoRA: Low-rank adaptation of large language models

    Edward J Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen-Zhu, Yuanzhi Li, Shean Wang, Lu Wang, and Weizhu Chen. LoRA: Low-rank adaptation of large language models. InInternational Conference on Learning Representations (ICLR), 2022

    2022

  14. [14]

    Rendering natural camera bokeh effect with deep learning

    Andrey Ignatov, Jagruti Patel, and Radu Timofte. Rendering natural camera bokeh effect with deep learning. InProceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR) Workshops, 2020

    2020

  15. [15]

    Kraus and M

    M. Kraus and M. Strengert. Depth-of-field rendering by pyramidal image processing.Computer Graphics Forum, 26(3):645–654, 2007

    2007

  16. [16]

    Flux.https://github.com/black-forest-labs/flux, 2024

    Black Forest Labs. Flux.https://github.com/black-forest-labs/flux, 2024

    2024

  17. [17]

    Iterative filter adaptive network for single image defocus deblurring

    Junyong Lee, Hyeongseok Son, Jaesung Rim, Sunghyun Cho, and Seungyong Lee. Iterative filter adaptive network for single image defocus deblurring. InProceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2021

    2021

  18. [18]

    Depth-of-field rendering with multiview synthesis

    Sungkil Lee, Elmar Eisemann, and Hans-Peter Seidel. Depth-of-field rendering with multiview synthesis. ACM Transactions on Graphics, 28(5):1–6, 2009

    2009

  19. [19]

    Real-time lens blur effects and focus control.ACM Transactions on Graphics, 29:65:1–65:7, 2010

    Sungkil Lee, Elmar Eisemann, and Hans-Peter Seidel. Real-time lens blur effects and focus control.ACM Transactions on Graphics, 29:65:1–65:7, 2010

    2010

  20. [20]

    Decoupled weight decay regularization

    Ilya Loshchilov and Frank Hutter. Decoupled weight decay regularization. InInternational Conference on Learning Representations (ICLR), 2019. 17

    2019

  21. [21]

    Dual pixel exploration: Simultaneous depth estimation and image restoration

    Liyuan Pan, Shah Chowdhury, Richard Hartley, Miaomiao Liu, Hongguang Zhang, and Hongdong Li. Dual pixel exploration: Simultaneous depth estimation and image restoration. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2021

    2021

  22. [22]

    Scalable diffusion models with transformers

    William Peebles and Saining Xie. Scalable diffusion models with transformers. InProceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), 2023

    2023

  23. [23]

    Interactive portrait bokeh rendering system

    Juewen Peng, Xianrui Luo, Ke Xian, and Zhiguo Cao. Interactive portrait bokeh rendering system. In IEEE International Conference on Image Processing (ICIP), 2021

    2021

  24. [24]

    BokehMe: When neural rendering meets classical rendering

    Juewen Peng, Zhiguo Cao, Xianrui Luo, Hao Lu, Ke Xian, and Jianming Zhang. BokehMe: When neural rendering meets classical rendering. InProceedings of the IEEE/CVF International Conference on Computer Vision and Pattern Recognition (CVPR), 2022

    2022

  25. [25]

    MPIB: An mpi-based bokeh rendering framework for realistic partial occlusion effects

    Juewen Peng, Jianming Zhang, Xianrui Luo, Hao Lu, Ke Xian, and Zhiguo Cao. MPIB: An mpi-based bokeh rendering framework for realistic partial occlusion effects. InEuropean conference on computer vision (ECCV), 2022

    2022

  26. [26]

    Juewen Peng, Zhiguo Cao, Xianrui Luo, Ke Xian, Wenfeng Tang, Jianming Zhang, and Guosheng Lin. BokehMe++: Harmonious fusion of classical and neural rendering for versatile bokeh creation.IEEE Transactions on Pattern Analysis and Machine Intelligence (PAMI), 47(3):1530–1547, 2025

    2025

  27. [27]

    A lens and aperture camera model for synthetic image generation

    Michael Potmesil and Indranil Chakravarty. A lens and aperture camera model for synthetic image generation. InInternational Conference on Computer Graphics and Interactive Techniques, 1981

    1981

  28. [28]

    Single image defocus deblurring via implicit neural inverse kernels

    Yuhui Quan, Xin Yao, and Hui Ji. Single image defocus deblurring via implicit neural inverse kernels. In Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), 2023

    2023

  29. [29]

    Learning to deblur using light field generated and real defocus images

    Lingyan Ruan, Bin Chen, Jizhou Li, and Miuling Lam. Learning to deblur using light field generated and real defocus images. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition (CVPR), 2022

    2022

  30. [30]

    Bokehlicious: Photorealistic bokeh rendering with controllable apertures

    Tim Seizinger, Florin-Alexandru Vasluianu, Marcos Conde, Zongwei Wu, and Radu Timofte. Bokehlicious: Photorealistic bokeh rendering with controllable apertures. InProceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), 2025

    2025

  31. [31]

    Yichen Sheng, Zixun Yu, Lu Ling, Zhiwen Cao, Xuaner Zhang, Xin Lu, Ke Xian, Haiting Lin, and Bedrich Benes. Dr. Bokeh: Differentiable occlusion-aware bokeh rendering. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2024

    2024

  32. [32]

    Prior and prediction inverse kernel transformer for single image defocus deblurring

    Peng Tang, Zhiqiang Xu, Chunlai Zhou, Pengfei Wei, Peng Han, Xin Cao, and Tobias Lasser. Prior and prediction inverse kernel transformer for single image defocus deblurring. InProceedings of the AAAI Conference on Artificial Intelligence, 2024

    2024

  33. [33]

    Generative refocusing: Flexible defocus control from a single image.arXiv preprint arXiv:2512.16923, 2025

    Chun-Wei Tuan Mu, Jia-Bin Huang, and Yu-Lun Liu. Generative refocusing: Flexible defocus control from a single image.arXiv preprint arXiv:2512.16923, 2025

    arXiv 2025

  34. [34]

    Jacobs, Bryan E

    Neal Wadhwa, Rahul Garg, David E. Jacobs, Bryan E. Feldman, Nori Kanazawa, Robert Carroll, Yair Movshovitz-Attias, Jonathan T. Barron, Yael Pritch, and Marc Levoy. Synthetic depth-of-field with a single-camera mobile phone.ACM Transactions on Graphics, 37(4):1–13, 2018

    2018

  35. [35]

    DiffCamera: Arbitrary refocusing on images

    Yiyang Wang, Xi Chen, Xiaogang Xu, Yu Liu, and Hengshuang Zhao. DiffCamera: Arbitrary refocusing on images. InProceedings of the SIGGRAPH Asia 2025 Conference Papers, 2025

    2025

  36. [36]

    Rendering realistic spectral bokeh due to lens stops and aberrations.The Visual Computer, 29, 2012

    Jiaze Wu, Changwen Zheng, Xiaohui Hu, and Fanjiang Xu. Rendering realistic spectral bokeh due to lens stops and aberrations.The Visual Computer, 29, 2012

    2012

  37. [37]

    K3DN: Disparity-aware kernel estimation for dual- pixel defocus deblurring

    Yan Yang, Liyuan Pan, Liu Liu, and Miaomiao Liu. K3DN: Disparity-aware kernel estimation for dual- pixel defocus deblurring. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2023

    2023

  38. [38]

    Real-time depth of field rendering via dynamic light field generation and filtering.Computer Graphics Forum, 29(7):2099–2107, 2010

    Xuan Yu, Rui Wang, and Jingyi Yu. Real-time depth of field rendering via dynamic light field generation and filtering.Computer Graphics Forum, 29(7):2099–2107, 2010

    2099

  39. [39]

    Generative photography: Scene-consistent camera control for realistic text-to-image synthesis.arXiv preprint arXiv: 2412.02168, 2024

    Yu Yuan, Xijun Wang, Yichen Sheng, Prateek Chennuri, Xingguang Zhang, and Stanley Chan. Generative photography: Scene-consistent camera control for realistic text-to-image synthesis.arXiv preprint arXiv: 2412.02168, 2024. 18

    arXiv 2024

  40. [40]

    Restormer: Efficient transformer for high-resolution image restoration

    Syed Waqas Zamir, Aditya Arora, Salman Khan, Munawar Hayat, Fahad Shahbaz Khan, and Ming-Hsuan Yang. Restormer: Efficient transformer for high-resolution image restoration. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition (CVPR), 2022

    2022

  41. [41]

    Efros, Eli Shechtman, and Oliver Wang

    Richard Zhang, Phillip Isola, Alexei A. Efros, Eli Shechtman, and Oliver Wang. The unreasonable effectiveness of deep features as a perceptual metric. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2018

    2018

  42. [42]

    Bilateral reference for high-resolution dichotomous image segmentation.CAAI Artificial Intelligence Research, 3:9150038, 2024

    Peng Zheng, Dehong Gao, Deng-Ping Fan, Li Liu, Jorma Laaksonen, Wanli Ouyang, and Nicu Sebe. Bilateral reference for high-resolution dichotomous image segmentation.CAAI Artificial Intelligence Research, 3:9150038, 2024

    2024

  43. [43]

    BokehDiff: Neural lens blur with one-step diffusion

    Chengxuan Zhu, Qingnan Fan, Qi Zhang, Jinwei Chen, Huaqi Zhang, Chao Xu, and Boxin Shi. BokehDiff: Neural lens blur with one-step diffusion. InProceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), 2025

    2025

  44. [44]

    Estimating spatially varying defocus blur from a single image.IEEE Transactions on Image Processing (TIP), 22(12):4879–4891, 2013

    Xiang Zhu, Scott Cohen, Stephen Schiller, and Peyman Milanfar. Estimating spatially varying defocus blur from a single image.IEEE Transactions on Image Processing (TIP), 22(12):4879–4891, 2013. 19

    2013