Pith. sign in

REVIEW 3 major objections 5 minor 57 references

Adaptive High-Pass Kernel Prediction for Efficient Video Deblurring

T0 review · 3 major / 5 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read A network that predicts adaptive high-pass kernels from Sobel and temporal gradient bases achieves 33.25 dB PSNR on GOPRO, the best among models with similar training memory, and runs 35x faster than ERDN.

desk verdict Solid low-budget video deblurring with a real 35x speedup, but the paper's theoretical guarantee for the high-pass kernel is unsupported and likely wrong as stated. read the letter →

arxiv 2412.01559 v1 pith:HJPTYG56 submitted 2024-12-02 cs.CV

Bo Ji , Angela Yao This is my paper · ORCID
classification cs.CV
keywords videodeblurringhigh-frequencyextractionadaptivekernelpredictionhigh-passfiltersefficientinferencespectralbiasnetworksGOPRO-DVDbenchmarks
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

Blur suppresses the high-frequency content of video, yet standard deep network components—self-attention and ordinary convolutions—tend to weaken or under-learn exactly those frequencies. The paper's thesis is that explicit high-frequency extraction is the missing ingredient, so it builds AHFNet around a module that predicts four mixing coefficients over a fixed set of high-pass basis kernels: two Sobel filters and two temporal differencing kernels. The generated kernel is convolved with the blurry input window, and the extracted feature maps feed a recurrent deblurring backbone, generalizing classical unsharp masking. On GOPRO the full model reaches 33.25 dB PSNR and 0.9439 SSIM, the best among models with comparable training memory, while requiring about 461 GMACs and 0.144 seconds per 1280×720 frame—roughly 35 times faster than ERDN. The authors report that the simplest version of the module, spatial Sobel plus temporal gradients, adds 0.33 dB PSNR over an equal-complexity control, which they take as evidence that the explicit high-pass structure, not extra capacity, drives the improvement.

What carries the argument

The adaptive high-frequency extraction module $\mathcal{H}$ is the load-bearing component. It holds $M=4$ fixed 3D high-pass basis kernels $\{\tilde{k}_j\}$—two normalized 3×3 Sobel filters (horizontal and vertical) and two temporal differencing kernels, one of which is Gram–Schmidt orthogonalized—and a small coefficient generator $G$ that reads the blurry input window and outputs mixing coefficients $\{\alpha_{t,j}\}$. The generated kernel $k_t=\sum_j \alpha_{t,j}\tilde{k}_j$ is convolved with the input, along with its 90°-rotated counterpart, and the outputs $h_t$, $h'_t$, and $|h_t|+|h'_t|$ form the extracted high-frequency feature set. Proposition 1 is the mechanism that connects this construction to the high-pass guarantee: any non-negative linear combination of high-pass filters is claimed to remain high-pass, which is why the network can predict only $M$ coefficients instead of the full kernel. The module is placed at the head of the network, with each of $N$ paths feeding a shared recurrent extraction module that produces the residual added to the input frame.

What would settle it

Evaluate the frequency response of a non-negative linear combination of two ideal high-pass filters with cutoffs at 0.1 and 0.9; under the standard half-power cutoff definition, the combined response reaches its half-power point above 0.9, which would directly falsify Proposition 1's 'no greater than the smallest cutoff' claim. A separate experiment could train AHFNet with a kernel that is provably high-pass by construction and compare against the current coefficient-prediction module; equal performance would show the guarantee is not what drives the accuracy.

Watch

Extended reading notes

Core claim

The central claim is that deblurring can be reformulated as explicit high-frequency recovery: a blurry frame $x_t$ plus a learned residual $\mathcal{L}(x_t, \mathcal{H}(\{x_i\}_{i=t-l}^{t+l}))$, where $\mathcal{H}$ is a module whose kernels are constrained to be high-pass. The constraint is enforced by construction: the module predicts non-negative mixing coefficients $\{\alpha_{t,j}\}_{j=1}^M$ and forms $k_t = \sum_j \alpha_{t,j} \tilde{k}_j$ from four fixed basis kernels—two normalized 3×3 Sobel kernels and two temporal difference kernels—then applies $k_t$, its 90°-rotated version, and the summed absolute values to the input window. Proposition 1 is invoked to guarantee that $k_t$ remains high-pass. The paper reports that this module lifts PSNR by 0.33 dB over a matched 'RGB×2' control, outperforms variants with randomly generated high-pass bases by 0.8 dB, and yields the best PSNR/SSIM among models trained within a comparable memory budget: 33.25 dB / 0.9439 on GOPRO and 33.19 dB / 0.9400 on DVD.

Load-bearing premise

The load-bearing premise is Proposition 1, stated in Section 3.3 without proof in the main text: any non-negative linear combination of high-pass filters is itself a high-pass filter with cutoff no greater than the smallest basis cutoff; if that frequency guarantee is wrong, the generated kernels are not high-pass by construction and the reported improvements would rest on the network learning useful filters on its own.

Editorial extensions

If this is right

  • If the central claim is right, the deblurring head does not need expensive transformer or deformable-convolution components: a handful of constrained convolutional paths suffices, cutting training memory to about one-sixth of models like FGST and inference to 0.144 s per frame.
  • The coefficient-prediction design reduces the learnable weights of the dynamic kernel from the full kernel size to $M=4$, so the module adds little in parameters or GMACs (461.35 vs 410.73 with no extraction paths) while improving PSNR by 0.33 dB.
  • Because the basis kernels are fixed and interpretable, the resulting model behaves as a learned generalization of unsharp masking, suggesting that frequency-based inductive biases can substitute for black-box architectural complexity in restoration tasks.
  • The module's gains are concentrated in mid-to-high frequency subbands, where the paper shows lower relative MSE than the 3D-Conv and Naive-Kernel variants, so the benefit is specifically about high-frequency recovery.
  • Scaling the number of extraction paths $N$ from 0 to 6 monotonically improves PSNR on GOPRO (32.56 to 32.89 dB), indicating the module is a composable building block.

Reading between the lines

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

  • An implication the paper leaves implicit is that the same module could transfer to single-image deblurring, super-resolution, or denoising, where spectral bias also suppresses fine details; nothing in the module is video-specific beyond the temporal basis kernels.
  • If Proposition 1's cutoff claim is not valid under the standard -3 dB definition, the empirical results would still stand, but the explanation would shift from an enforced spectral guarantee to the inductive bias of fixed gradient bases—a distinction future ablations could test by comparing against a provably high-pass kernel of the same capacity.
  • A testable extension would be spatially varying coefficients: the current module predicts one mixing coefficient per kernel per frame, so the adaptive kernel is global; allowing per-pixel or per-patch coefficients could adapt the filter to local blur anisotropy at a modest computational cost.
  • The success of coefficient prediction over four bases hints that larger basis sets, such as oriented edge filters at multiple scales, could extend the same efficiency-accuracy trade-off, since only the coefficient dimension grows.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 5 minor

Summary. The paper proposes AHFNet, a video deblurring network that explicitly extracts high-frequency features by dynamically predicting mixing coefficients over a fixed set of high-pass basis kernels (spatial Sobel filters and temporal difference kernels). The generated kernels are convolved with the input video, along with their 90-degree rotations, and the resulting features are fed into a recurrent deblurring module. The authors report state-of-the-art performance among methods with similar training memory footprint on GOPRO and DVD, with 33.25 dB PSNR on GOPRO, 461 GMACs, and 0.144 s inference time, and they attribute the gains to the explicit high-pass structure, supported by Proposition 1.

Significance. If the results hold, AHFNet offers a practical efficiency/performance trade-off for video deblurring, with a simple and interpretable design that is substantially cheaper in training memory than transformer/deformable-convolution baselines while achieving competitive PSNR/SSIM. Strengths include the clear architectural description, the frequency-subband analysis in Figure 6, ablations over the number of extraction paths (Table 4) and basis kernel choices (Table 5), and the promise of released code. The main caveats are the theoretical guarantee (Proposition 1) and the statistical robustness of the small performance differences.

major comments (3)
  1. [Section 3.3, Proposition 1 and Eq. (4)-(5)] The proposition's claim that the cutoff frequency of a non-negative linear combination of high-pass filters is not greater than the smallest basis cutoff fc1 is false under the standard -3 dB definition. For two ideal high-pass filters with cutoffs 0.1 and 0.9, the combined response is 0 below 0.1, 1 between 0.1 and 0.9, and 2 above 0.9; normalized to the passband maximum of 2, the -3 dB crossing occurs at 0.9, not at or below 0.1. The proof is deferred to supplementary material and is not available in this version. Furthermore, Proposition 1 requires non-negative coefficients αi, but Eq. (5) defines α via an unconstrained network G, and no non-negativity activation (e.g., ReLU or softmax) is described in the architecture or training details. Without non-negativity, the generated kernel kt need not have zero DC response, so even the weak 'zero DC gain' notion of high-pass is not guaranteed. This undermines the central interpretability claim that AHFNet explicitly constrains the network to high-frequency extraction. Please correct the proposition or its proof, enforce or justify non-negativity of α, and/or directly measure the frequency responses of learned kernels to empirically support the high-pass claim.
  2. [Section 4.4, Table 5 and Abstract] The headline improvement of 0.33 dB PSNR from 'integrating spatial and temporal gradients' is computed against the '+ RGB×2' control, which lacks any high-pass basis and simply repeats the early module with RGB features. The appropriate ablation for isolating the effect of the high-pass constraint is the comparison against the equally complex '3D Conv' and 'Naive Kernels' variants in Table 3, where the gains are 0.07 dB and 0.09 dB, respectively. As reported, the abstract and Section 4.4 overstate the contribution of the explicit high-frequency structure, conflating the basis-kernel choice with the overall module design. Please report the comparison against the learned-kernel variant as the primary ablation or revise the claim accordingly.
  3. [Tables 1-5] No variance or number of runs is reported for PSNR/SSIM, and several key differences are small (e.g., Table 3: AHFNet 32.89 vs 3DConv 32.82; Table 4: N=4 vs N=6, 0.18 dB). Without error bars or significance tests, the claimed improvements over the closest baselines are not statistically supported. Please report mean and standard deviation over at least three runs, or otherwise demonstrate that the differences exceed run-to-run variability.
minor comments (5)
  1. [Section 4.1] There is a stray period and line break after 'including both forward and backward propagation.' in the memory footprint definition; please remove the extra punctuation.
  2. [Section 3.3] The notation for the temporal basis kernels is ambiguous: the first kernel is written as [[0],[−1],[1]] and the second as [[1],[−1/2],[−1/2]]; please clarify the frame ordering (e.g., t−1, t, t+1) and the Gram-Schmidt derivation, which appears to start from [[1],[−1],[0]] rather than from the first basis kernel.
  3. [Section 4.4] The symbols ∇x, ∇2x, and ∇tx are used without formal definition; please define them (e.g., first-order spatial gradient, Laplacian, temporal gradient) in the text or in a table.
  4. [Figure 6] The y-axis label 'Relative MSE Values' and the subband labels are difficult to read; consider enlarging the font and using a clearer axis description.
  5. [Section 4.2, Figure 3] The claim of 'optimal trade-off' is based on a single point per method; consider adding error bars or noting that memory values are taken from official implementations.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the empirical claims are evaluated on held-out test data and the theoretical claim, while possibly unsupported, is not circular.

full rationale

The paper's central empirical result, AHFNet's PSNR/SSIM on GOPRO and DVD, is produced by training on standard Charbonnier loss and evaluated on held-out test frames; no reported number is derived from a fitted constant or from the definition of the module. The choice of Sobel and temporal-gradient basis kernels is guided by validation comparisons in Table 5, which is design tuning rather than a prediction, and the final result is still measured on test data. Ablations against 3D Conv, Naive Kernels, and RGB-x-k controls provide independent evidence for the contribution of the high-pass module. The only self-citation (reference [12]) appears in related work and is not load-bearing for any derivation. Proposition 1 in Section 3.3 is load-bearing for the interpretability claim that the generated kernel is guaranteed high-pass, but the issue with it is one of correctness and support (the proof is deferred to supplementary, the cutoff claim is questionable under the -3 dB definition, and the non-negative coefficient assumption is not reported as enforced), not circularity: the proposition is a mathematical assertion about basis kernels, not a restatement of the empirical outputs. Accordingly, no circular step can be exhibited, and the paper is not circular in the sense defined here.

Assumptions & free parameters 2 free parameters · 3 assumptions · 0 invented entities

The central claim is empirical and no physical constant is fitted. The hand-designed basis kernels and N are the main tunable choices that affect the reported numbers; these are validated on the test set, which introduces a mild selection-bias risk.

free parameters (2)
  • basis kernel set = Sobel horizontal/vertical and two temporal differencing kernels
    Handpicked in Sec 3.3, then validated in Sec 4.4 against Kirsch, Laplacian, random filters. The chosen set maximizes PSNR in Table 5, so the reported result depends on this selection.
  • number of extraction paths N = 6
    Table 4 shows higher N gives higher PSNR; N=6 was adopted for the final model. This is a validated hyperparameter.
assumptions (3)
  • domain assumption Video blur model: blurry frame equals g((1/B) sum of sharp frames) (Eq 1)
    Borrowed from [25] to justify viewing blur as low-pass filtering. Not tested in this paper.
  • domain assumption Self-attention and standard convolutions suppress high frequencies; networks show spectral bias
    Based on cited works [31,33,40]; motivates the explicit high-pass module but not demonstrated by the paper's own analysis.
  • ad hoc to paper Proposition 1: a non-negative linear combination of high-pass filters is high-pass with cutoff not greater than the smallest input cutoff (Eq 4)
    The main text states it without proof, deferring to supplementary. The cutoff-frequency claim is questionable under the standard -3 dB definition, and the paper relies on this proposition to guarantee the generated kernel remains high-pass.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Adaptive High-Pass Kernel Prediction for Efficient Video Deblurring." pith.science (2026). https://pith.science/paper/HJPTYG56

@misc{pith2026241201559,
  author       = {Pith},
  title        = {Pith review of: Adaptive High-Pass Kernel Prediction for Efficient Video Deblurring},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/HJPTYG56}},
  note         = {Machine review of arXiv:2412.01559}
}
read the original abstract

State-of-the-art video deblurring methods use deep network architectures to recover sharpened video frames. Blurring especially degrades high-frequency (HF) information, yet this aspect is often overlooked by recent models that focus more on enhancing architectural design. Recovering these fine details is challenging, partly due to the spectral bias of neural networks, which are inclined towards learning low-frequency functions. To address this, we enforce explicit network structures to capture the fine details and edges. We dynamically predict adaptive high-pass kernels from a linear combination of high-pass basis kernels to extract high-frequency features. This strategy is highly efficient, resulting in low-memory footprints for training and fast run times for inference, all while achieving state-of-the-art when compared to low-budget models. The code is available at https://github.com/jibo27/AHFNet.

Figures

Figures reproduced from arXiv: 2412.01559 by the authors.

Figure 1
Figure 1. Adaptive high-frequency extraction module ( [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Overview of AHFNet. We extract high frequencies using [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Trade-off between PSNR on GOPRO and the training [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: Inference time for a single 1280 × 720 frame vs. PSNR on the GOPRO dataset [25]. The dot size represents GMACs. Compared models share a similar training memory footprint. Our method achieves the best PSNR with minimal inference time [PITH_FULL_IMAGE:figures/full_fig_p…
Figure 5
Figure 5. Figure 5: Qualitative comparisons to models with a similar training memory footprint. [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]
Figure 6
Figure 6. Figure 6: Comparison of learning in various frequency subbands . [PITH_FULL_IMAGE:figures/full_fig_p007_6.png]
Figure 7
Figure 7. Figure 7: Examples of learned kernels and features on GOPRO dataset (please zoom in for a better view). [PITH_FULL_IMAGE:figures/full_fig_p008_7.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

57 extracted references · 47 canonical work pages

  1. [1]

    Compute-efficient deep learning: Algorithmic trends and opportunities

    Brian R Bartoldson, Bhavya Kailkhura, and Davis Blalock. Compute-efficient deep learning: Algorithmic trends and opportunities. Journal of Machine Learning Research , 24(122):1–77, 2023. 1

    work page 2023

  2. [2]

    Blind motion deblurring using multiple images

    Jian-Feng Cai, Hui Ji, Chaoqiang Liu, and Zuowei Shen. Blind motion deblurring using multiple images. Journal of computational physics, 228(14):5057–5071, 2009. 1, 2

    work page 2009

  3. [3]

    Basicvsr: The search for essential com- ponents in video super-resolution and beyond

    Kelvin CK Chan, Xintao Wang, Ke Yu, Chao Dong, and Chen Change Loy. Basicvsr: The search for essential com- ponents in video super-resolution and beyond. In Proceed- ings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 4947–4956, 2021. 4

    work page 2021

  4. [4]

    Two deterministic half-quadratic regular- ization algorithms for computed imaging

    Pierre Charbonnier, Laure Blanc-Feraud, Gilles Aubert, and Michel Barlaud. Two deterministic half-quadratic regular- ization algorithms for computed imaging. In Proceedings of 1st international conference on image processing, volume 2, pages 168–172. IEEE, 1994. 5

    work page 1994

  5. [5]

    Dynamic convolu- tion: Attention over convolution kernels

    Yinpeng Chen, Xiyang Dai, Mengchen Liu, Dongdong Chen, Lu Yuan, and Zicheng Liu. Dynamic convolu- tion: Attention over convolution kernels. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 11030–11039, 2020. 2

    work page 2020

  6. [6]

    Fast motion deblurring

    Sunghyun Cho and Seungyong Lee. Fast motion deblurring. In ACM SIGGRAPH Asia 2009 papers, pages 1–8. 2009. 2

    work page 2009

  7. [7]

    A generalized unsharp masking algorithm

    Guang Deng. A generalized unsharp masking algorithm. IEEE transactions on Image Processing , 20(5):1249–1261,

    work page

  8. [8]

    Digital image process- ing

    Rafael C Gonzales and Paul Wintz. Digital image process- ing. Addison-Wesley Longman Publishing Co., Inc., 1987. 1

    work page 1987

Show all 57 references
  1. [9]

    Memory-efficient deformable convolution based joint denoising and demosaic- ing for uhd images

    Juntao Guan, Rui Lai, Yang Lu, Yangang Li, Huanan Li, Lichen Feng, Yintang Yang, and Lin Gu. Memory-efficient deformable convolution based joint denoising and demosaic- ing for uhd images. IEEE Transactions on Circuits and Sys- tems for Video Technology, 32(11):7346–7358, 2022. 2

    2022

  2. [10]

    Guided image fil- tering

    Kaiming He, Jian Sun, and Xiaoou Tang. Guided image fil- tering. IEEE transactions on pattern analysis and machine intelligence, 35(6):1397–1409, 2012. 2

    2012

  3. [11]

    Digital image processing

    Bernd J ¨ahne. Digital image processing. Springer Science & Business Media, 2005. 2

    2005

  4. [12]

    Multi-scale memory-based video deblurring

    Bo Ji and Angela Yao. Multi-scale memory-based video deblurring. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages 1919– 1928, 2022. 2, 6

    1919

  5. [13]

    Xydeblur: divide and conquer for single image deblurring

    Seo-Won Ji, Jeongmin Lee, Seung-Wook Kim, Jun-Pyo Hong, Seung-Jin Baek, Seung-Won Jung, and Sung-Jea Ko. Xydeblur: divide and conquer for single image deblurring. In Proceedings of the IEEE/CVF conference on computer vi- sion and pattern recognition, pages 17421–17430, 2022. 2

    2022

  6. [14]

    Dynamic filter networks

    Xu Jia, Bert De Brabandere, Tinne Tuytelaars, and Luc V Gool. Dynamic filter networks. Advances in neural informa- tion processing systems, 29, 2016. 2, 4, 6

    2016

  7. [15]

    Erdn: Equivalent receptive field deformable network for video deblurring

    Bangrui Jiang, Zhihuai Xie, Zhen Xia, Songnan Li, and Shan Liu. Erdn: Equivalent receptive field deformable network for video deblurring. In Computer Vision–ECCV 2022: 17th European Conference, Tel Aviv, Israel, October 23–27, 2022, Proceedings, Part XVIII, pages 663–678. Spri...

    2022

  8. [16]

    Adam: A method for stochastic optimization

    Diederik P Kingma and Jimmy Ba. Adam: A method for stochastic optimization. arXiv preprint arXiv:1412.6980 ,

    arXiv

  9. [17]

    Blind de- convolution using a normalized sparsity measure

    Dilip Krishnan, Terence Tay, and Rob Fergus. Blind de- convolution using a normalized sparsity measure. In CVPR 2011, pages 233–240. IEEE, 2011. 1, 2

    2011

  10. [18]

    A simple baseline for video restoration with grouped spatial- temporal shift

    Dasong Li, Xiaoyu Shi, Yi Zhang, Ka Chun Cheung, Simon See, Xiaogang Wang, Hongwei Qin, and Hongsheng Li. A simple baseline for video restoration with grouped spatial- temporal shift. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages ...

    2023

  11. [19]

    Arvo: Learning all-range volumetric correspondence for video deblurring

    Dongxu Li, Chenchen Xu, Kaihao Zhang, Xin Yu, Yiran Zhong, Wenqi Ren, Hanna Suominen, and Hongdong Li. Arvo: Learning all-range volumetric correspondence for video deblurring. In Proceedings of the IEEE/CVF Con- ference on Computer Vision and Pattern Recognition, pages 7721–77...

    2021

  12. [20]

    Recurrent video restoration trans- former with guided deformable attention

    Jingyun Liang, Yuchen Fan, Xiaoyu Xiang, Rakesh Ranjan, Eddy Ilg, Simon Green, Jiezhang Cao, Kai Zhang, Radu Timofte, and Luc V Gool. Recurrent video restoration trans- former with guided deformable attention. Advances in Neu- ral Information Processing Systems, 35:378–393, 20...

    2022

  13. [21]

    Enhanced deep residual networks for single image super-resolution

    Bee Lim, Sanghyun Son, Heewon Kim, Seungjun Nah, and Kyoung Mu Lee. Enhanced deep residual networks for single image super-resolution. In Proceedings of the IEEE confer- ence on computer vision and pattern recognition workshops, pages 136–144, 2017. 5

    2017

  14. [22]

    Flow-guided sparse transformer for video deblurring

    Jing Lin, Yuanhao Cai, Xiaowan Hu, Haoqian Wang, You- liang Yan, Xueyi Zou, Henghui Ding, Yulun Zhang, Radu Timofte, and Luc Van Gool. Flow-guided sparse transformer for video deblurring. arXiv preprint arXiv:2201.01893 ,

    arXiv

  15. [23]

    Sgdr: Stochas- tic gradient descent with warm restarts

    Ilya Loshchilov and Frank Hutter. Sgdr: Stochas- tic gradient descent with warm restarts. arXiv preprint arXiv:1608.03983, 2016. 5

    2016 arXiv

  16. [24]

    Weightnet: Revisiting the design space of weight networks

    Ningning Ma, Xiangyu Zhang, Jiawei Huang, and Jian Sun. Weightnet: Revisiting the design space of weight networks. In European Conference on Computer Vision , pages 776–

  17. [25]

    Deep multi-scale convolutional neural network for dynamic scene deblurring

    Seungjun Nah, Tae Hyun Kim, and Kyoung Mu Lee. Deep multi-scale convolutional neural network for dynamic scene deblurring. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 3883–3891,

  18. [26]

    Re- current neural networks with intra-frame iterations for video deblurring

    Seungjun Nah, Sanghyun Son, and Kyoung Mu Lee. Re- current neural networks with intra-frame iterations for video deblurring. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 8102–8111,

  19. [27]

    Cascaded deep video deblurring using temporal sharpness prior

    Jinshan Pan, Haoran Bai, and Jinhui Tang. Cascaded deep video deblurring using temporal sharpness prior. InProceed- ings of the IEEE/CVF conference on computer vision and pattern recognition, pages 3043–3051, 2020. 1, 2, 5, 6, 7 9

    2020

  20. [28]

    IEEE transactions on pattern analysis and machine intelligence, 39(2):342–355, 2016

    Jinshan Pan, Zhe Hu, Zhixun Su, and Ming-Hsuan Yang.l 0- regularized intensity and gradient prior for deblurring text images and beyond. IEEE transactions on pattern analysis and machine intelligence, 39(2):342–355, 2016. 1, 2

    2016

  21. [29]

    Blind image deblurring using dark channel prior

    Jinshan Pan, Deqing Sun, Hanspeter Pfister, and Ming- Hsuan Yang. Blind image deblurring using dark channel prior. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 1628–1636, 2016. 2

    2016

  22. [30]

    Deep discriminative spatial and temporal net- work for efficient video deblurring

    Jinshan Pan, Boming Xu, Jiangxin Dong, Jianjun Ge, and Jinhui Tang. Deep discriminative spatial and temporal net- work for efficient video deblurring. In The IEEE Confer- ence on Computer Vision and Pattern Recognition(CVPR) , Feb 2023. 1

    2023

  23. [31]

    How do vision transformers work? arXiv preprint arXiv:2202.06709, 2022

    Namuk Park and Songkuk Kim. How do vision transformers work? arXiv preprint arXiv:2202.06709, 2022. 1, 3

    2022 arXiv

  24. [32]

    Image enhancement via adaptive unsharp masking

    Andrea Polesel, Giovanni Ramponi, and V John Mathews. Image enhancement via adaptive unsharp masking. IEEE transactions on image processing, 9(3):505–510, 2000. 2

    2000

  25. [33]

    On the spectral bias of neural networks

    Nasim Rahaman, Aristide Baratin, Devansh Arpit, Felix Draxler, Min Lin, Fred Hamprecht, Yoshua Bengio, and Aaron Courville. On the spectral bias of neural networks. In International Conference on Machine Learning , pages 5301–5310. PMLR, 2019. 1

  26. [34]

    Optical flow estima- tion using a spatial pyramid network

    Anurag Ranjan and Michael J Black. Optical flow estima- tion using a spatial pyramid network. In Proceedings of the IEEE conference on computer vision and pattern recogni- tion, pages 4161–4170, 2017. 5

    2017

  27. [35]

    Two motion-blurred images are better than one

    Alex Rav-Acha and Shmuel Peleg. Two motion-blurred images are better than one. Pattern recognition letters , 26(3):311–317, 2005. 2

    2005

  28. [36]

    High-quality mo- tion deblurring from a single image

    Qi Shan, Jiaya Jia, and Aseem Agarwala. High-quality mo- tion deblurring from a single image. Acm transactions on graphics (tog), 27(3):1–10, 2008. 2

    2008

  29. [37]

    Real-time single image and video super-resolution using an efficient sub-pixel convolutional neural network

    Wenzhe Shi, Jose Caballero, Ferenc Husz ´ar, Johannes Totz, Andrew P Aitken, Rob Bishop, Daniel Rueckert, and Zehan Wang. Real-time single image and video super-resolution using an efficient sub-pixel convolutional neural network. In Proceedings of the IEEE conference on compu...

    2016

  30. [38]

    Deep video deblurring for hand-held cameras

    Shuochen Su, Mauricio Delbracio, Jue Wang, Guillermo Sapiro, Wolfgang Heidrich, and Oliver Wang. Deep video deblurring for hand-held cameras. In Proceedings of the IEEE conference on computer vision and pattern recogni- tion, pages 1279–1288, 2017. 2, 5, 6, 7

    2017

  31. [39]

    Edge-based blur kernel estimation using patch priors

    Libin Sun, Sunghyun Cho, Jue Wang, and James Hays. Edge-based blur kernel estimation using patch priors. In IEEE international conference on computational photogra- phy (ICCP), pages 1–8. IEEE, 2013. 2

    2013

  32. [40]

    Defects of convolutional decoder networks in frequency representation

    Ling Tang, Wen Shen, Zhanpeng Zhou, Yuefeng Chen, and Quanshi Zhang. Defects of convolutional decoder networks in frequency representation. arXiv preprint arXiv:2210.09020, 2022. 1, 3

    2022 arXiv

  33. [41]

    Scale-recurrent network for deep image deblurring

    Xin Tao, Hongyun Gao, Xiaoyong Shen, Jue Wang, and Ji- aya Jia. Scale-recurrent network for deep image deblurring. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 8174–8182, 2018. 6

    2018

  34. [42]

    Anti-oversmoothing in deep vision trans- formers via the fourier domain analysis: From theory to practice

    Peihao Wang, Wenqing Zheng, Tianlong Chen, and Zhangyang Wang. Anti-oversmoothing in deep vision trans- formers via the fourier domain analysis: From theory to practice. arXiv preprint arXiv:2203.05962, 2022. 1, 3, 8

    2022 arXiv

  35. [43]

    Edvr: Video restoration with enhanced deformable convolutional networks

    Xintao Wang, Kelvin CK Chan, Ke Yu, Chao Dong, and Chen Change Loy. Edvr: Video restoration with enhanced deformable convolutional networks. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, pages 0–0, 2019. 1, 2, 5

    2019

  36. [44]

    Efficient video deblurring guided by motion magnitude

    Yusheng Wang, Yunfan Lu, Ye Gao, Lin Wang, Zhihang Zhong, Yinqiang Zheng, and Atsushi Yamashita. Efficient video deblurring guided by motion magnitude. In Computer Vision–ECCV 2022: 17th European Conference, Tel Aviv, Israel, October 23–27, 2022, Proceedings, Part XIX , pages ...

    2022

  37. [45]

    Image quality assessment: from error visibility to structural similarity

    Zhou Wang, Alan C Bovik, Hamid R Sheikh, and Eero P Si- moncelli. Image quality assessment: from error visibility to structural similarity. IEEE transactions on image processing, 13(4):600–612, 2004. 5

    2004

  38. [46]

    Basis prediction networks for effective burst denoising with large kernels

    Zhihao Xia, Federico Perazzi, Micha ¨el Gharbi, Kalyan Sunkavalli, and Ayan Chakrabarti. Basis prediction networks for effective burst denoising with large kernels. In Proceed- ings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 11844–11853, 2020. 2, 6

    2020

  39. [47]

    Two-phase kernel estimation for robust motion deblurring

    Li Xu and Jiaya Jia. Two-phase kernel estimation for robust motion deblurring. In Computer Vision–ECCV 2010: 11th European Conference on Computer Vision, Heraklion, Crete, Greece, September 5-11, 2010, Proceedings, Part I 11, pages 157–170. Springer, 2010. 2

    2010

  40. [48]

    Unnatural l0 sparse representation for natural image deblurring

    Li Xu, Shicheng Zheng, and Jiaya Jia. Unnatural l0 sparse representation for natural image deblurring. In Proceed- ings of the IEEE conference on computer vision and pattern recognition, pages 1107–1114, 2013. 1, 2

    2013

  41. [49]

    A variational em framework with adaptive edge selection for blind motion deblurring

    Liuge Yang and Hui Ji. A variational em framework with adaptive edge selection for blind motion deblurring. In Pro- ceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 10167–10176, 2019. 2

    2019

  42. [50]

    Blurriness-guided un- sharp masking

    Wei Ye and Kai-Kuang Ma. Blurriness-guided un- sharp masking. IEEE Transactions on Image Processing , 27(9):4465–4477, 2018. 1, 2

    2018

  43. [51]

    Multi- image blind deblurring using a coupled adaptive sparse prior

    Haichao Zhang, David Wipf, and Yanning Zhang. Multi- image blind deblurring using a coupled adaptive sparse prior. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pages 1051–1058, 2013. 1, 2

    2013

  44. [52]

    Spatio- temporal deformable attention network for video deblurring

    Huicong Zhang, Haozhe Xie, and Hongxun Yao. Spatio- temporal deformable attention network for video deblurring. In Computer Vision–ECCV 2022: 17th European Confer- ence, Tel Aviv, Israel, October 23–27, 2022, Proceedings, Part XVI, pages 581–596. Springer, 2022. 6, 7

    2022

  45. [53]

    Pixel screening based intermediate correction for blind de- blurring

    Meina Zhang, Yingying Fang, Guoxi Ni, and Tieyong Zeng. Pixel screening based intermediate correction for blind de- blurring. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages 5892– 5900, 2022. 2

    2022

  46. [54]

    Kbnet: Kernel basis network for image restoration

    Yi Zhang, Dasong Li, Xiaoyu Shi, Dailan He, Kangning Song, Xiaogang Wang, Hongwei Qin, and Hongsheng Li. Kbnet: Kernel basis network for image restoration. arXiv preprint arXiv:2303.02881, 2023. 2 10

    2023 arXiv

  47. [55]

    Efficient spatio-temporal recurrent neural network for video deblurring

    Zhihang Zhong, Ye Gao, Yinqiang Zheng, and Bo Zheng. Efficient spatio-temporal recurrent neural network for video deblurring. In Computer Vision–ECCV 2020: 16th Euro- pean Conference, Glasgow, UK, August 23–28, 2020, Pro- ceedings, Part VI 16, pages 191–207. Springer, 2020. 2, 6, 7

    2020

  48. [56]

    Spatio-temporal filter adaptive network for video deblurring

    Shangchen Zhou, Jiawei Zhang, Jinshan Pan, Haozhe Xie, Wangmeng Zuo, and Jimmy Ren. Spatio-temporal filter adaptive network for video deblurring. In Proceedings of the IEEE/CVF international conference on computer vision, pages 2482–2491, 2019. 2

    2019

  49. [57]

    Decon- volving psfs for a better motion deblurring using multiple images

    Xiang Zhu, Filip ˇSroubek, and Peyman Milanfar. Decon- volving psfs for a better motion deblurring using multiple images. In Computer Vision–ECCV 2012: 12th European Conference on Computer Vision, Florence, Italy, October 7- 13, 2012, Proceedings, Part V 12, pages 636–647. Springer,

    2012

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.