AsyncDSB: Schedule-Asynchronous Diffusion Schr\"odinger Bridge for Image Inpainting

Baoquan Zhang; Guangming Ye; Guotao Liang; Kenghong Lin; Lisai Zhang; Shanshan Feng; Xiaochen Qi; Yunming Ye; Zihao Han

REVIEW 5 major objections 6 minor 39 references

AsyncDSB: Schedule-Asynchronous Diffusion Schr\"odinger Bridge for Image Inpainting

T0 review · 5 major / 6 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read Existing diffusion Schrödinger bridge inpainting applies one noise schedule to all pixels, but AsyncDSB assigns each pixel its own shifted schedule based on gradient magnitude, improving FID by about 3% to 14% over I2SB.

desk verdict Novel per-pixel asynchronous schedule idea that works empirically, but the bridge validity is not proven and the evaluation needs error bars. read the letter →

arxiv 2412.08149 v1 pith:TVBDZJVK submitted 2024-12-11 cs.CV

Zihao Han , Baoquan Zhang , Lisai Zhang , Shanshan Feng , Kenghong Lin , Guotao Liang , Yunming Ye , Xiaochen Qi

show 1 more author

Guangming Ye

This is my paper · ORCID

classification cs.CV

keywords imageinpaintingdiffusionSchrödingerbridgenoiseschedulepixel-asynchronousgradientpriorCelebA-HQPlaces2

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper argues that diffusion Schrödinger bridge inpainting, as instantiated by I2SB, is suboptimal because it applies the same noise schedule $\beta_t$ to every pixel while real restoration is asynchronous: high-frequency pixels (edges, contours) come together before low-frequency pixels (flat color, texture fill). It proposes AsyncDSB, which first predicts the gradient of the masked region, converts gradient magnitude into a per-pixel time shift $\tau_{i,j}$, and then runs the bridge with a pixel-specific schedule $\beta_{t,i,j}$. On CelebA-HQ and Places2, this raises FID by roughly 3% to 14% over I2SB, with the biggest gains on masks that destroy structure. If correct, the result shows that aligning the noise schedule to the per-pixel order of restoration is a direct and cheap lever for inpainting quality.

What carries the argument

The central object is the pixel-asynchronous noise schedule $\beta_{t,i,j}$, built by shifting a global schedule per pixel through a predicted-gradient time shift $\tau_{i,j}$ (Eq. 11). The shift map comes from a gradient completion network $f_{\theta_g}$ trained adversarially with feature matching, following the edge-prediction approach of EdgeConnect. Each pixel then has its own accumulated variances $\bar{\sigma}_{t,i,j}^2$ and $\sigma_{t,i,j}^2$ (Eq. 12), and inference uses the pixel-wise posterior sampling recursion of Eq. (13). The mechanism works by allocating early steps to high-gradient pixels and later steps to low-gradient pixels, explicitly modeling the temporal ordering observed in real restoration.

What would settle it

Train the same AsyncDSB pipeline but replace the gradient-derived $\tau$ map with a random or shuffled per-pixel shift that has the same distribution and no correlation with image structure. If FID remains at the same level, the improvement comes from adding per-pixel variation rather than from restoring high-frequency pixels first. Alternatively, directly measure per-pixel restoration times in AsyncDSB trajectories; the mechanism is refuted if high-frequency pixels are not restored earlier than low-frequency pixels on average.

Watch

Extended reading notes

Core claim

The paper's central discovery is a schedule-restoration mismatch in I2SB. Using the analytic Gaussian posterior of the Schrödinger bridge, it derives that the restoration speed of a pixel is $\partial x_t/\partial t = -\beta_t (x_0 \odot x_m)/(\bar{\sigma}_t^2 + \sigma_t^2)$, so $\beta_t$ sets the global pace while the pixel value sets the amplitude. Measuring the derivative of SSIM across denoising time on CelebA-HQ shows the practical restoration curve lags the theoretical one, and the lag is concentrated in mid- and low-frequency pixels, meaning pixels are not restored synchronously. The paper's fix is to give each pixel its own time shift $\tau_{i,j}$, computed by normalizing a Gaussian-filtered predicted gradient map to $[\tau_{\min}, \tau_{\max}]$, and to run posterior sampling with a per-pixel schedule. This re-aligns practical and theoretical restoration speeds and improves FID.

Load-bearing premise

All of the reported gains rest on the assumption that a score network trained with the global, synchronous schedule still gives correct per-pixel denoising predictions when each pixel is advanced at its own shifted time; the paper does not prove this, it only demonstrates the empirical improvement.

Editorial extensions

If this is right

AsyncDSB improves FID by roughly 3% to 14% over I2SB on CelebA-HQ and Places2 across center, half, wide, and narrow masks, with larger gains on masks that remove structure.
The asynchronous schedule visibly narrows the gap between theoretical and practical restoration-speed curves, particularly for mid- and low-frequency pixels.
The framework is not limited to gradient guidance; the authors state that any $\beta_{t,i,j}$ strategy matching the pixel-asynchronous generation process can be plugged in.
For higher mask ratios, the optimal schedule shifts earlier (larger $\tau_{\min}$ and $\tau_{\max}$), indicating that mask-aware schedule selection is a promising extension the paper leaves for future work.

Reading between the lines

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

The same per-pixel schedule idea could transfer to other bridge-based image translation tasks, such as super-resolution or deblurring, where restoration order is also spatially non-uniform.
A cheaper proxy for the gradient map — e.g., distance to the mask boundary or local variance — might capture most of the gain, which would separate the benefit of asynchrony from the specific gradient predictor.
The paper does not prove that per-pixel shifted schedules preserve the Schrödinger bridge coupling or that the score network stays calibrated for pixels at different effective times; if that proof fails, the empirical gains would still stand but the theoretical explanation would need revision.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

Let me give you my read of AsyncDSB. The core idea is new and worth knowing: instead of one global noise schedule for all pixels, each pixel gets its own schedule shifted by a tau_{i,j} derived from predicted gradient magnitude. The motivation is grounded in an observation that high-frequency pixels restore earlier than low-frequency ones during I2SB sampling (their Figure 3). And they do run a controlled ablation: adding the gradient map as a UNet condition to I2SB (I2SB+gradient) does not reproduce the gains, so the improvement is specifically from the asynchronous schedule, not just the extra conditioning.

Where it gets soft is the theory. The stress-test note is right: they train the score network on a global beta_t schedule, then at inference feed the network xt at global time t while using per-pixel variances from a different local schedule. The posterior sampling in Eq. (13) is only the I2SB posterior if the pixel's noise level equals the training-time noise level at that t. They never prove that, and nothing in the paper corrects for the mismatch. That means the per-pixel update is a heuristic, not a tested bridge. It may work anyway—empirically it does—but the paper frames it more formally than it is.

The empirical support is also thinner than the headline suggests. FID gains over I2SB are 3-14% depending on mask and dataset, several around 3-4%, with no error bars or significance tests (the checklist admits that). And tau_min/tau_max are tuned per mask class on the evaluation set; Figure 6 shows sensitivity. Code is not released.

Still, I don't see a load-bearing flaw. The mechanism is transparent, the ablation is fair, and the gains are consistent. The paper overclaims theoretical backing, but the empirical finding is plausible and the asynchronous-schedule idea could transfer to other image-to-image tasks.

Who it's for: anyone working on diffusion bridges or schedule design for restoration. I'd give it a serious referee—the idea is novel and the experiments are enough to warrant scrutiny. But it needs a revision that either proves the per-pixel posterior validity or reframes the method as a heuristic and adds error bars.

Referee Report

5 major / 6 minor

Summary. The paper analyzes the noise schedule of I2SB and argues that a single global schedule beta_t is mismatched to the asynchronous way image pixels are restored. It proposes AsyncDSB, which (i) completes the gradient map of the corrupted region with an adversarially trained module, (ii) converts gradient magnitudes into per-pixel shift parameters tau_{i,j} via Eq. (11), and (iii) applies per-pixel posterior sampling with pixel-dependent variances sigma_{t,i,j}. Experiments on CelebA-HQ and Places2 report FID improvements over I2SB, with relative gains of roughly 3% to 14% depending on mask type.

Significance. If correct, the idea that per-pixel asynchronous schedules can better align the theoretical and practical restoration processes in diffusion Schrodinger bridges is a useful and non-obvious contribution to image inpainting. The paper identifies a real phenomenon: the empirical restoration speed of I2SB lags the theoretical schedule, and this lag differs across frequency bands. The proposed fix is simple, practical, and testable, and the authors include a useful control (I2SB plus gradient condition) to separate the effect of the asynchronous schedule from the effect of the extra gradient input. The main strength is the clean diagnosis of restoration asynchrony and the straightforward gradient-prior schedule. However, the paper does not yet establish that the per-pixel shifted schedule is a valid Schrodinger bridge, and the reported gains are not accompanied by measures of statistical uncertainty.

major comments (5)

[Step 2, Eqs. (11)-(13)] The per-pixel noise schedule beta_{t,i,j} is never explicitly defined as a function of t and tau_{i,j}. The text says that tau_{i,j} transforms the traditional noise schedule into a pixel-wise form, but no equation gives beta_{t,i,j} in terms of beta_t and tau_{i,j}. This is needed both for reproducibility and for checking the validity of the bridge. More importantly, Eq. (13) evaluates the score as s_theta(x_{t,i,j}, t)/sigma_t using the global sigma_t, while the variances in Eq. (12) are per-pixel sigma_{t,i,j}. For the posterior update to be a valid discretization of a Schrodinger bridge, the score must correspond to the per-pixel marginal q(x_{t,i,j} | x0, x1) at the per-pixel noise level Sigma_{t,i,j}; using a score trained on the global schedule on a mixture of pixels at different effective noise levels is not justified. The authors need to either prove that the per-pixel time shifts preserve the score calibration, retrain or condition the score on per-pixel noise levels, or restrict the theoretical claims to the empirical heuristic they actually implement.
[Theoretical Analysis, Eq. (7)] The derivation of restoration speed drops the stochastic Term3 with the statement that its expectation is zero. For individual trajectories, the noise term contributes to the per-pixel restoration order, and Eq. (11) assigns schedules to individual pixels based on gradient magnitude alone. An expectation-based argument therefore does not establish that high-gradient pixels restore earlier along a typical sample path. If the intended claim is about the expected restoration speed, the paper should say so explicitly and relate the measured SSIM derivative in Figure 2 to the expectation rather than to individual pixel behavior.
[Experiments, Table 1 and Reproducibility Checklist] The headline FID improvements are reported without error bars, seeds, or statistical tests; the reproducibility checklist explicitly answers 'no' to the use of statistical tests. Many absolute FID differences are small (e.g., 2.2 to 1.9 on CelebA-HQ center, 8.6 to 8.3 on Places2 center) and could lie within run-to-run variation. Additionally, tau_min and tau_max are tuned per mask type and dataset (Figure 6, Table 2), with no validation split described. The authors should provide multiple seeds with means and variances, and a significance test for the I2SB comparison, especially because the 3%-14% relative improvement is the central empirical claim.
[Ablation Study, Table 3] The I2SB+gradient baseline in Table 3 is not described in sufficient detail: it is unclear how the gradient map is injected into the UNet, whether the same gradient completion network is used, and whether the baseline is trained with the same objective as AsyncDSB. Without these details, Table 3 cannot cleanly attribute the improvement to the asynchronous schedule rather than to the way the gradient conditioning is implemented.
[Figure 7 and Section 'Can AsyncDSB solve the schedule-restoration mismatch issue?'] The claim that AsyncDSB alleviates the schedule-restoration mismatch is, in part, enforced by design: Eq. (11) assigns earlier schedules to pixels with larger gradients, so observing in Figure 7 that high-frequency pixels restore earlier than low-frequency ones is not an independent confirmation of the mechanism. The FID improvements in Table 1 are the relevant empirical evidence; the text should not present Figure 7 as a verification of the scheduling hypothesis without acknowledging that the schedule directly encodes that hypothesis.

minor comments (6)

[Preliminaries, Eq. (2)] The notation in Eq. (2) is ambiguous because both variances are rendered as sigma^2_t; based on Eq. (6), the first should be sigma-bar^2_t and the second sigma^2_t. Please use distinct symbols consistently.
[Theoretical Analysis and Abstract] Several cross-references are missing or blank, such as 'Section' and 'our insight of Section' in the Introduction and Methodology sections. Please fill in the appropriate section numbers.
[Table 1] The abstract and text state a '3% - 14% improvement' over I2SB, but some entries in Table 1 report gains in the ‰ range (e.g., IS 6.71 to 6.74) while others are in the % range. The reporting should use a consistent relative measure and avoid mixing percentages with permille.
[Figure 6] The figure would be easier to interpret with explicit axis labels for tau_min and tau_max, a colorbar with units, and a legend or note explaining that blank cells are invalid because tau_min must not exceed tau_max.
[Reproducibility Checklist] The checklist states that all source code is included in an appendix and will be made public, but no code appendix appears in the manuscript. Either include the code or remove these claims.
[Eq. (13)] The integral in Eq. (13) is typeset incorrectly; it should be written as an integral from t - 1/T to t, not as 'Z t/(t-1/T)'. Please fix the notation.

Circularity Check

1 steps flagged · score 4.0 of 10

The schedule-restoration alignment claim is partially enforced by construction, but the headline FID gains are externally benchmarked.

self definitional [Methodology, Step 2, Eq. (11); Experiments, 'Can AsyncDSB solve the schedule-restoration mismatch issue?' (Fig. 7)]
"Intuitively, the pixel with high frequency should assign larger τi,j than pixel with low frequency since the pixel with high frequency are priority restored. ... τi,j = gauss(ˆxgg i,j)−min(gauss(ˆxgg)) / max(gauss(ˆxgg))−min(gauss(ˆxgg)) (τmax−τmin) +τmin ... From results, we can see that the schedule-restoration mismatch issue can be significantly alleviated after applying our AsyncDSB."

Eq. (11) constructs τ_{i,j} directly from the gradient/frequency map, with the stated rationale that high-frequency pixels should restore earlier. Because β_{t,i,j} and hence σ²_{t,i,j}, σ̄²_{t,i,j} are per-pixel functions of τ (Eq. 12), and posterior sampling (Eq. 13) uses those per-pixel schedules, the ordering of effective restoration times across pixels is imposed by construction. Therefore Fig. 7's finding that after applying AsyncDSB the practical restoration process matches the theoretical schedule and that high-frequency pixels restore earlier is not an independent empirical confirmation of the asynchronous-restoration hypothesis; it reads back the τ assignment.

full rationale

The paper's central quantitative claim—roughly 3%–14% FID improvement over I2SB—is an external benchmark comparison and is not derived from the schedule equations, so it does not reduce by construction. No load-bearing self-citation chain or imported uniqueness theorem is present. The circular component is narrower but real: the asynchronous schedule is defined from the same gradient/frequency ordering that Fig. 7 then 'verifies.' Eq. (11) assigns each pixel a shifted schedule τ_{i,j} from the predicted gradient map, Eq. (12) computes per-pixel variances from that schedule, and Eq. (13) samples with those per-pixel schedules; consequently, high-frequency pixels restoring earlier is a designed property, not a learned or predicted outcome. The mismatch-alleviation evidence in Fig. 7 therefore loops back to the construction. The appended reproducibility checklist also states that no statistical significance tests were used ('The significance of any improvement ... is judged using appropriate statistical tests. (no)'), and Fig. 6 shows τ_min/τ_max selected per mask on the evaluated benchmark; these are robustness/overfitting concerns rather than definitional circularity, but they reinforce caution about the strength of the headline improvements. Overall, partial circularity in the mechanistic claim, independent external grounding for the main FID result, and no problematic self-citation yield a score of 4.

Assumptions & free parameters 4 free parameters · 5 assumptions · 0 invented entities

No new physical or mathematical entities are proposed; the per-pixel schedule is a modification of existing noise schedule hyperparameters. The free parameters are the tuning knobs that carry much of the method's empirical success.

free parameters (4)

tau_min = 0.2T for center masks; 0.001T for wide/narrow; 0.001T to 0.7T by mask ratio in Table 2
Tuned per mask type and mask ratio to minimize FID (Figure 6, Table 2); no principled derivation is given.
tau_max = 0.5T for center masks; 0.4T for wide masks; 0.9T in mask-ratio table
Tuned jointly with tau_min; controls the asynchronous spread of the schedule.
Gaussian filter parameters in gauss() = not reported
The kernel size and sigma for smoothing the gradient map are unspecified but directly affect the schedule map in Eq(11).
lambda_adv and lambda_FM = not specified in main text
Weights in the gradient completion loss Eq(10); hyperparameters from EdgeConnect-style adversarial training.

assumptions (5)

standard math I2SB analytic Gaussian posterior and score matching objective
Eq(2) to Eq(4) from Liu et al. 2023 are used without re-derivation; the paper inherits their validity.
domain assumption Mask corruption model x1 = (1 - xm) * x0
Used in Eq(8) to derive restoration speed; assumes corrupted pixels are exactly zeroed out.
ad hoc to paper High-frequency-first restoration priority
Core heuristic: pixels with large gradients should be scheduled earlier; motivated by observation, not proved necessary.
ad hoc to paper Per-pixel shifted schedule remains a valid Schrödinger bridge
Eq(12) and Eq(13) apply pixel-specific schedules without a proof that the bridge coupling or score calibration is preserved.
domain assumption Accuracy of the predicted gradient map
The schedule relies on the gradient completion network f_theta_g; if the predicted gradients are wrong, the schedule misassigns priorities.

how reviews work

0 comments

Cite this review

Pith. "Pith review of AsyncDSB: Schedule-Asynchronous Diffusion Schr\"odinger Bridge for Image Inpainting." pith.science (2026). https://pith.science/paper/TVBDZJVK

@misc{pith2026241208149,
  author       = {Pith},
  title        = {Pith review of: AsyncDSB: Schedule-Asynchronous Diffusion Schr\"odinger Bridge for Image Inpainting},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/TVBDZJVK}},
  note         = {Machine review of arXiv:2412.08149}
}

read the original abstract

Image inpainting is an important image generation task, which aims to restore corrupted image from partial visible area. Recently, diffusion Schr\"odinger bridge methods effectively tackle this task by modeling the translation between corrupted and target images as a diffusion Schr\"odinger bridge process along a noising schedule path. Although these methods have shown superior performance, in this paper, we find that 1) existing methods suffer from a schedule-restoration mismatching issue, i.e., the theoretical schedule and practical restoration processes usually exist a large discrepancy, which theoretically results in the schedule not fully leveraged for restoring images; and 2) the key reason causing such issue is that the restoration process of all pixels are actually asynchronous but existing methods set a synchronous noise schedule to them, i.e., all pixels shares the same noise schedule. To this end, we propose a schedule-Asynchronous Diffusion Schr\"odinger Bridge (AsyncDSB) for image inpainting. Our insight is preferentially scheduling pixels with high frequency (i.e., large gradients) and then low frequency (i.e., small gradients). Based on this insight, given a corrupted image, we first train a network to predict its gradient map in corrupted area. Then, we regard the predicted image gradient as prior and design a simple yet effective pixel-asynchronous noise schedule strategy to enhance the diffusion Schr\"odinger bridge. Thanks to the asynchronous schedule at pixels, the temporal interdependence of restoration process between pixels can be fully characterized for high-quality image inpainting. Experiments on real-world datasets show that our AsyncDSB achieves superior performance, especially on FID with around 3% - 14% improvement over state-of-the-art baseline methods.

Figures

Figures reproduced from arXiv: 2412.08149 by the authors.

**Figure 1.** Connection between noise schedule (βt) and the restoration process of I2SB(Liu et al. 2023). noise schedule actually mismatch the asynchronous pixels restoration process; and 3) Because of the above setting mismatch, the existing I2SB method (Liu et al. 2023) suffers from a schedule-restoration mismatching issue, i.e., there is usually a large discrepancy between the theoretical denoise schedule and practical resto… view at source ↗

**Figure 2.** Qualitative Analysis. theoretical schedule; but 2) the overall tread has been delayed by a large margin. This means that existing I2SB methods suffers from a schedule-restoration mismatching issue, i.e., there is a large discrepancy between theoretical schedule (see blue line) and practical restoration process (see blue dotted line). Such mismatching issue will results in that the theoretical nonlinear schedule is … view at source ↗

**Figure 3.** Comparison between theoretical schedule and prac [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗

Figures from the paper (4 more)

**Figure 4.** Figure 4: The overall framework of our AsyncDSB, which consists of two steps, an image gradient completion step, and a [PITH_FULL_IMAGE:figures/full_fig_p005_4.png]

**Figure 5.** Figure 5: Qualitative results on CelebA-HQ and Places2 under regular and irregular masks. [PITH_FULL_IMAGE:figures/full_fig_p006_5.png]

**Figure 6.** Figure 6: Analysis of τ min and τ max on Places2. 0T 1.0T × 10−4 0 -0.6 -1.2 t 𝐼𝐼2𝑆𝑆𝑆𝑆’s Theoretical 𝐼𝐼2𝑆𝑆𝑆𝑆’s Practical Ours Theoretical Ours Practical 0.04T0.08T 0T 1.0T × 10−4 0 -0.6 -1.2 t 0.06T0.21T 0T 1.0T × 10−4 0 -0.6 -1.2 t 0.06T0.31T Restoration Speed high-frequency mi…

**Figure 8.** Figure 8: Sampling process of our AsyncDSB. Can AsyncDSB solve the schedule-restoration mismatch issue? In [PITH_FULL_IMAGE:figures/full_fig_p007_8.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

39 extracted references · 28 canonical work pages

[1]

, " * write output.state after.block = add.period write newline

ENTRY address archivePrefix author booktitle chapter edition editor eid eprint howpublished institution isbn journal key month note number organization pages publisher school series title type volume year label extra.label sort.label short.list INTEGERS output.state before.all mid.sentence after.sentence after.block FUNCTION init.state.consts #0 'before.a...
[2]

write newline

" write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION word.in bbl.in capitalize " " * FUNCT...
[3]

Ballester, C.; Bertalmio, M.; Caselles, V.; Sapiro, G.; and Verdera, J. 2001. Filling-in by joint interpolation of vector fields and gray levels. IEEE transactions on image processing, 10(8): 1200--1211

work page 2001
[4]

Barnes, C.; Shechtman, E.; Finkelstein, A.; and Goldman, D. B. 2009. PatchMatch: A randomized correspondence algorithm for structural image editing. ACM Trans. Graph., 28(3): 24

2009
[5]

Bendel, M.; Ahmad, R.; and Schniter, P. 2023. A Regularized Conditional GAN for Posterior Sampling in Image Recovery Problems. Advances in Neural Information Processing Systems, 36

work page 2023
[6]

Bertalmio, M.; Sapiro, G.; Caselles, V.; and Ballester, C. 2000. Image inpainting. In Proceedings of the 27th annual conference on Computer graphics and interactive techniques, 417--424

2000
[7]

Bertalmio, M.; Vese, L.; Sapiro, G.; and Osher, S. 2003. Simultaneous structure and texture image inpainting. IEEE transactions on image processing, 12(8): 882--889

work page 2003
[8]

Chen, T. 2023. On the importance of noise scheduling for diffusion models. arXiv preprint arXiv:2301.10972

arXiv 2023

Show all 39 references

[9]

T.; and Pavon, M

Chen, Y.; Georgiou, T. T.; and Pavon, M. 2016. On the relation between optimal transport and Schr \"o dinger bridges: A stochastic control viewpoint. Journal of Optimization Theory and Applications, 169: 671--691

2016
[10]

Chu, T.; Chen, J.; Sun, J.; Lian, S.; Wang, Z.; Zuo, Z.; Zhao, L.; Xing, W.; and Lu, D. 2023. Rethinking fast fourier convolution in image inpainting. In Proceedings of the IEEE/CVF International Conference on Computer Vision, 23195--23205

2023
[11]

Criminisi, A.; Perez, P.; and Toyama, K. 2003. Object removal by exemplar-based inpainting. In 2003 IEEE Computer Society Conference on Computer Vision and Pattern Recognition, 2003. Proceedings., volume 2, II--II. IEEE

2003
[12]

Criminisi, A.; P \'e rez, P.; and Toyama, K. 2004. Region filling and object removal by exemplar-based image inpainting. IEEE Transactions on image processing, 13(9): 1200--1212

2004
[13]

De Bortoli, V.; Thornton, J.; Heng, J.; and Doucet, A. 2021. Diffusion schr \"o dinger bridge with applications to score-based generative modeling. Advances in Neural Information Processing Systems, 34: 17695--17709

2021
[14]

Dong, Q.; Cao, C.; and Fu, Y. 2022. Incremental transformer structure enhanced image inpainting with masking positional encoding. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 11358--11368

2022
[15]

Feng, T.; Feng, W.; Li, W.; and Lin, D. 2022. Cross-image context for single image inpainting. Advances in Neural Information Processing Systems, 35: 1474--1487

2022
[16]

Guo, Q.; Liu, S.; Yu, Y.; and Luo, P. 2023. Rethinking the Noise Schedule of Diffusion-Based Generative Models

2023
[17]

Ho, J.; Jain, A.; and Abbeel, P. 2020. Denoising diffusion probabilistic models. Advances in neural information processing systems, 33: 6840--6851

2020
[18]

Li, B.; Xue, K.; Liu, B.; and Lai, Y.-K. 2023. Bbdm: Image-to-image translation with brownian bridge diffusion models. In Proceedings of the IEEE/CVF conference on computer vision and pattern Recognition, 1952--1961

2023
[19]

Li, W.; Lin, Z.; Zhou, K.; Qi, L.; Wang, Y.; and Jia, J. 2022 a . Mat: Mask-aware transformer for large hole image inpainting. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 10758--10768

2022
[20]

Li, X.; Guo, Q.; Lin, D.; Li, P.; Feng, W.; and Wang, S. 2022 b . Misf: Multi-level interactive siamese filtering for high-fidelity image inpainting. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 1869--1878

2022
[21]

Lin, S.; Liu, B.; Li, J.; and Yang, X. 2024. Common diffusion noise schedules and sample steps are flawed. In Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision, 5404--5411

2024
[22]

A.; Nie, W.; and Anandkumar, A

Liu, G.-H.; Vahdat, A.; Huang, D.-A.; Theodorou, E. A.; Nie, W.; and Anandkumar, A. 2023. I2SB: image-to-image Schr \"o dinger bridge. In Proceedings of the 40th International Conference on Machine Learning, 22042--22062

2023
[23]

Liu, H.; Wang, Y.; Qian, B.; Wang, M.; and Rui, Y. 2024. Structure Matters: Tackling the Semantic Discrepancy in Diffusion Models for Image Inpainting. arXiv preprint arXiv:2403.19898

2024 arXiv
[24]

Liu, X.; and Wu, L. 2023. Learning diffusion bridges on constrained domains. In international conference on learning representations (ICLR)

2023
[25]

Liu, Z.; Luo, P.; Wang, X.; and Tang, X. 2015. Deep Learning Face Attributes in the Wild. In Proceedings of International Conference on Computer Vision (ICCV)

2015
[26]

Lugmayr, A.; Danelljan, M.; Romero, A.; Yu, F.; Timofte, R.; and Van Gool, L. 2022. Repaint: Inpainting using denoising diffusion probabilistic models. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 11461--11471

2022
[27]

Z.; and Ebrahimi, M

Nazeri, K.; Ng, E.; Joseph, T.; Qureshi, F. Z.; and Ebrahimi, M. 2019. Edgeconnect: Generative image inpainting with adversarial edge learning. arXiv preprint arXiv:1901.00212

2019 arXiv
[28]

Q.; and Dhariwal, P

Nichol, A. Q.; and Dhariwal, P. 2021. Improved denoising diffusion probabilistic models. In International conference on machine learning, 8162--8171. PMLR

2021
[29]

Pavon, M. 2002. Quantum Schr \"o dinger bridges. In Directions in mathematical systems theory and optimization, 227--238. Springer

2002
[30]

Peluchetti, S. 2023. Diffusion bridge mixture transports, Schr \"o dinger bridge problems and generative modeling. Journal of Machine Learning Research, 24(374): 1--51

2023
[31]

Sargsyan, A.; Navasardyan, S.; Xu, X.; and Shi, H. 2023. Mi-gan: A simple baseline for image inpainting on mobile devices. In Proceedings of the IEEE/CVF International Conference on Computer Vision, 7335--7345

2023
[32]

Shi, Y.; De Bortoli, V.; Campbell, A.; and Doucet, A. 2023. Diffusion Schr \"o dinger bridge matching. Advances in Neural Information Processing Systems, 36

2023
[33]

L.; and Lemaire, V

Strasman, S.; Ocello, A.; Boyer, C.; Corff, S. L.; and Lemaire, V. 2024. An analysis of the noise schedule for score-based generative models. arXiv preprint arXiv:2402.04650

2024 arXiv
[34]

Su, X.; Song, J.; Meng, C.; and Ermon, S. 2022. Dual Diffusion Implicit Bridges for Image-to-Image Translation. In The Eleventh International Conference on Learning Representations

2022
[35]

Wang, W.; Niu, L.; Zhang, J.; Yang, X.; and Zhang, L. 2022. Dual-path image inpainting with auxiliary gan inversion. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 11421--11430

2022
[36]

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

2023
[37]

Zeng, Y.; Lin, Z.; Lu, H.; and Patel, V. M. 2021. Cr-fill: Generative image inpainting with auxiliary contextual reconstruction. In Proceedings of the IEEE/CVF international conference on computer vision, 14164--14173

2021
[38]

S.; and Chang, S

Zhang, G.; Ji, J.; Zhang, Y.; Yu, M.; Jaakkola, T. S.; and Chang, S. 2023. Towards coherent image inpainting using denoising diffusion implicit models

2023
[39]

Zomet. 2003. Learning how to inpaint from global image statistics. In Proceedings Ninth IEEE international conference on computer vision, 305--312. IEEE

2003

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.

[1] [1]

, " * write output.state after.block = add.period write newline

ENTRY address archivePrefix author booktitle chapter edition editor eid eprint howpublished institution isbn journal key month note number organization pages publisher school series title type volume year label extra.label sort.label short.list INTEGERS output.state before.all mid.sentence after.sentence after.block FUNCTION init.state.consts #0 'before.a...

[2] [2]

write newline

" write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION word.in bbl.in capitalize " " * FUNCT...

[3] [3]

Ballester, C.; Bertalmio, M.; Caselles, V.; Sapiro, G.; and Verdera, J. 2001. Filling-in by joint interpolation of vector fields and gray levels. IEEE transactions on image processing, 10(8): 1200--1211

work page 2001

[4] [4]

Barnes, C.; Shechtman, E.; Finkelstein, A.; and Goldman, D. B. 2009. PatchMatch: A randomized correspondence algorithm for structural image editing. ACM Trans. Graph., 28(3): 24

2009

[5] [5]

Bendel, M.; Ahmad, R.; and Schniter, P. 2023. A Regularized Conditional GAN for Posterior Sampling in Image Recovery Problems. Advances in Neural Information Processing Systems, 36

work page 2023

[6] [6]

Bertalmio, M.; Sapiro, G.; Caselles, V.; and Ballester, C. 2000. Image inpainting. In Proceedings of the 27th annual conference on Computer graphics and interactive techniques, 417--424

2000

[7] [7]

Bertalmio, M.; Vese, L.; Sapiro, G.; and Osher, S. 2003. Simultaneous structure and texture image inpainting. IEEE transactions on image processing, 12(8): 882--889

work page 2003

[8] [8]

Chen, T. 2023. On the importance of noise scheduling for diffusion models. arXiv preprint arXiv:2301.10972

arXiv 2023