Pith. sign in

REVIEW 4 major objections 5 minor 40 references

DM$^3$Net: Dual-Camera Super-Resolution via Domain Modulation and Multi-scale Matching

T0 review · 4 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read A dual-camera super-resolution network that matches telephoto details at three scales exceeds prior methods on three real-world benchmark datasets.

desk verdict Solid incremental dual-camera SR paper whose corner-region gains need one more validation step, but the core matching + domain modulation stack is sound and worth referee time. read the letter →

arxiv 2506.06993 v1 pith:EUCBFLEK submitted 2025-06-08 cs.CV

Cong Guan , Jiacheng Ying , Yuya Ieiri , Osamu Yoshie This is my paper
classification cs.CV
keywords dual-camerasuper-resolutionreference-basedmulti-scalematchingdomainmodulationkeypruningopticalflowreal-worlddatasets
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

DM3Net is a dual-camera super-resolution method that uses a telephoto image as reference to sharpen a wide-angle image. The paper claims it achieves the best reported PSNR, SSIM, and LPIPS on three real-world datasets, in both the overlapping center region and the non-overlapping corners, and that it generalizes across datasets better than prior dual-camera and reference-based super-resolution methods. The core ideas are multi-scale patch matching, which retrieves reference details at three receptive-field scales, and domain modulation, which learns a compressed global representation of the gap between high-resolution and degraded imagery and uses it to condition reconstruction. A key-pruning step reduces memory and inference time with only a small accuracy loss.

What carries the argument

The central mechanism is the multi-scale matching module, which patchifies query and key features at three receptive fields (1/4, 1/2, and full scale), retrieves for each query patch the key patch with maximum cosine similarity, builds a matching index map and confidence map, and stitches the corresponding warped reference features with confidence weighting. Key Pruning accelerates this by sampling a subset of key patches, deleting those whose similarity to a sampled patch exceeds a threshold, and matching only against the survivors. The second mechanism is domain modulation: two Global Prior Extractors compress the LR-center/reference pair and the LR/ground-truth pair into 1024-dimensional embeddings, an L1 domain-aware loss aligns them, and the reconstruction module uses the embedding to predict affine parameters for feature modulation.

What would settle it

Take a real dual-camera pair whose wide-angle corner contains a unique high-frequency texture (such as a printed sign or fabric pattern) that does not appear anywhere in the overlapping field of view, and evaluate DM3Net's corner-region PSNR/SSIM against a single-image super-resolution baseline; if the corner metric no longer improves, the central assumption that the center contains transferable matches for corners is falsified.

Watch

Extended reading notes

Core claim

The central claim is that matching reference patches at multiple scales, rather than at a single coarse scale, makes detail transfer from the telephoto image reliable enough to reconstruct even the non-overlapping corner regions of the wide-angle image. In DM3Net, multi-scale query features from the low-resolution image are matched against key features from the low-resolution center crop at 1/4, 1/2, and full scale; for each query patch the most similar key patch is selected by cosine similarity, and the corresponding high-resolution reference features, aligned by RAFT optical flow, are reassembled and fused across scales. In parallel, two global prior extractors produce compressed domain embeddings, one from the LR-center/reference pair and one from the LR/ground-truth pair; a domain-aware loss aligns their distributions so that the reference-derived embedding can stand in for the ground-truth-derived one at inference time. These embeddings supply affine modulation parameters to the reconstruction module. On DuSR-Real, RealMCVSR-Real, and CameraFusion-Real, the authors report consistent gains over DCSR, KeDuSR, SelfDZSR, TTSR, MASA-SR, SwinIR, and Real-ESRGAN on full-image, center, and corner metrics, plus the best cross-dataset generalization when trained on one dataset and tested on the other two.

Load-bearing premise

The method assumes that every low-resolution patch, including those in the non-overlapping corners, has a sufficiently similar patch in the overlapping center region whose telephoto features can be transferred, and that the optical-flow and VGG-feature matching keep those retrieved patches correctly aligned.

Editorial extensions

If this is right

  • Multi-scale matching improves detail reconstruction in both the overlapping center and the non-overlapping corners of the wide-angle image, with full-scale matching contributing the largest single improvement.
  • Key Pruning cuts inference time from 31.08s to 11.32s and memory usage from 49.41GB to 40.95GB on CameraFusion-Real, at a cost of 0.12dB PSNR and 0.072 SSIM.
  • Models trained on DuSR-Real generalize to RealMCVSR-Real and CameraFusion-Real better than the compared methods, indicating the matching and embedding mechanisms are not tightly bound to a single dataset's degradations.
  • Adding the ground-truth-derived embedding zgt alongside the reference-derived embedding z further improves reconstruction, and the domain-aware loss that aligns z to zgt is what lets the network rely on z alone during inference.

Reading between the lines

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

  • We infer that the method's improvement over KeDuSR comes largely from matching at finer scales; a direct comparison that replaces DM3Net's multi-scale retrieval with a single full-scale retrieval while keeping all other components equal would isolate this contribution.
  • We infer that the domain-aware embedding could be estimated from the telephoto pair alone without ground truth, which would remove the need for HR ground truth during training and is a testable extension of the paper's domain-gap idea.
  • The paper's stated limitation — 27.42M parameters and roughly 1.4 seconds per 1792x896 output — suggests real-time smartphone deployment would require distillation or a lighter matching scheme; the authors say they plan to reduce parameters and increase speed.
  • An untested stress case is a wide-angle image whose corner contains a unique texture absent from the overlapping field of view; multiscale matching would then have no good source patch, and the method would likely degrade toward single-image super-resolution.
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

4 major / 5 minor

Summary. The paper proposes DM3Net, a dual-camera super-resolution network that combines global domain modulation with multi-scale patch matching. Two domain-aware embeddings are learned from LR-GT and LR-center/Ref pairs, and a multi-scale Q-K matching module using VGG features retrieves telephoto feature patches to enhance reconstruction, with a Key Pruning strategy to reduce cost. Experiments on DuSR-Real, RealMCVSR-Real, and CameraFusion-Real report state-of-the-art PSNR/SSIM/LPIPS results, and an appendix provides additional loss-variant comparisons, ablations, and efficiency analyses.

Significance. If the results are reproducible, DM3Net would be a meaningful advance for practical dual-camera SR, with consistent quantitative gains across three public benchmarks and a useful efficiency trade-off. The paper's strengths include the use of public datasets, retraining of baselines under the same training set, ablation studies isolating each contribution, and a cross-dataset generalization evaluation. The main limitations are the absence of uncertainty quantification for the headline numbers and the unvalidated assumption that LR-space similarity implies transferable telephoto detail for corner regions; both are fixable within a revision and should be addressed before the central SOTA claim is fully accepted.

major comments (4)
  1. [§3.3, Eqs. (3)–(4)] The corner-region advantage rests on the assumption that a high cosine similarity between a Q patch from the full LR image and a K patch from the LR center in VGG-19 feature space is a reliable indicator that the corresponding V patch (telephoto feature) contains transferable high-frequency detail for the queried patch. The confidence map is computed from Q-K similarity, not from any measure of whether the retrieved V patch improves reconstruction. The ablations in Table 5 show that adding scales helps, but they do not validate the correctness of the retrieved correspondences. To support the claimed corner-region gains (e.g., Table 4: 27.08 vs 26.43 PSNR for CameraFusion-Real), the authors should provide an oracle-correspondence check—for example, comparing against ground-truth flow or manually annotated correspondences, or corrupting retrieved patches and measuring the effect—or otherwise demonstrate that the retrieval is semantically correct.
  2. [§4.2, Tables 2–4] The central claim that DM3Net 'consistently achieves the best performance in all metrics' is based on single-run point estimates with no error bars, confidence intervals, or significance tests. Some advantages are small (e.g., Table 2: LPIPS 0.159 vs 0.164 for KeDuSR), and the reported numbers could fall within run-to-run variability. Please report results over multiple training runs (or at least bootstrapped confidence intervals over test images) and state the number of runs used; this is necessary to substantiate the SOTA claim.
  3. [Appendix B] The -ℓ comparison is not apples-to-apples. The text states that DM3Net-ℓ is trained with both the Charbonnier loss and the domain-aware loss Ldomain (Eq. 9, weight λ3=1000), while the baseline -ℓ models are trained with only L1/Charbonnier loss. The better metrics of DM3Net-ℓ in Tables 8–10 may therefore be due to the extra loss rather than the architectural contributions. Please retrain the baselines with the same loss terms (or the same total objective as DM3Net-ℓ), or explicitly justify why the extra loss does not affect the comparison. In addition, the claim that DM3Net-ℓ 'consistently achieves the best performance across most evaluation metrics' is not supported in Table 8, where KeDuSR-ℓ has higher center-region PSNR (29.58 vs 29.50).
  4. [Appendix C.1 and Table 1] The hyperparameters of Key Pruning (sampling interval 16, threshold 0.7) are selected based on curves of PSNR and inference time computed directly on the DuSR-Real dataset (Figure 7), and the same dataset is then used for the final quantitative results. This constitutes tuning on the test set and may make the reported efficiency-performance trade-off and the Table 1 numbers optimistic. The authors should perform hyperparameter selection on a validation split or a separate dataset and report the test-set results accordingly.
minor comments (5)
  1. [§4.3 and §4.5] There are typos in the text: 'SelfZeDuSR' in Section 4.3 should be 'SelfDZSR', and 'DUSR-Real' in Section 4.5 should be 'DuSR-Real'.
  2. [§4.1 and Table 1] Section 4.1 says all experiments are conducted on a single NVIDIA A100 40G GPU, but Table 1 states tests are on NVIDIA H20; please clarify which hardware was used for which experiments.
  3. [Tables 2 and 11] The latency values for TTSR and MASA-SR are swapped between Table 2 (TTSR 6.013, MASA-SR 7.067) and Table 11 (TTSR 7.067, MASA-SR 6.013); please correct the inconsistency.
  4. [Related Work and Section 4.1] ZeDuSR [22] is cited in the related work but is not included in the quantitative comparisons; please state whether it is applicable to this setting and, if so, include it or explain its exclusion.
  5. [Appendix A, Eq. (10)] Equation (10) contains a notational inconsistency: 'Ltotal = L_rec + λ3 Ldomain = λ1 Lrec + λ2 Lper + λ3 Ldomain' mixes L_rec and Lrec; please define Lrec consistently and align the notation.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity found: the paper's claims are empirical architecture results evaluated on public benchmarks, with no derivation step that reduces to its own inputs.

full rationale

The paper proposes a dual-camera super-resolution network and validates it experimentally. No load-bearing step equates an output with an input by construction. The domain-aware embeddings z and zgt are trained with the domain-aware loss Ldomain = ||z - zgt||_1, but this is a training objective used to make z mimic zgt; at inference only z is used, and z is not defined in terms of the evaluation metrics. The multi-scale matching module in Eqs. (3)-(4) selects reference patches by cosine similarity in VGG feature space; this is a design choice, not a circular reduction, and its accuracy is tested against external benchmarks. The key-pruning threshold and sampling interval are tuned on DuSR-Real as hyperparameters, but the final claims are not forced by that tuning because the model is also evaluated across three datasets with retrained baselines. Appendix B notes that the DM3Net-l variant receives an additional Ldomain loss while baseline -l variants do not; this is an unequal-comparison concern affecting fairness of that specific table, but it does not make the central result equivalent to its inputs. There are no self-citations carrying the argument, no imported uniqueness theorems, and no renamed known result presented as a derivation. The paper is self-contained: its contributions are architectural and empirical, with claims supported by external benchmark evaluations.

Assumptions & free parameters 4 free parameters · 5 assumptions · 2 invented entities

The central claim is an empirical benchmark result, not a derivation, so the ledger mostly records domain assumptions and hand-set hyperparameters. The only new latent constructs are the two domain embeddings; neither has an independent falsifiable handle beyond the paper's own ablations.

free parameters (4)
  • Domain-aware loss weight lambda_3 = 1000
    Set empirically in Appendix A; directly controls how strongly the inference embedding z is pulled toward the ground-truth embedding z_gt.
  • Perceptual loss weight lambda_2 = 0.01
    Set empirically in Appendix A; affects visual quality balance but not the core geometry.
  • Key pruning threshold = 0.7
    Chosen by balancing PSNR and inference time in Appendix C.1; higher thresholds retain more keys and cost more time.
  • Key pruning sampling interval = 16 (horizontal and vertical)
    Chosen in Appendix C.1; smaller intervals sample more key patches and increase computation.
assumptions (5)
  • domain assumption The three public datasets (DuSR-Real, RealMCVSR-Real, CameraFusion-Real) are representative of real dual-camera captures and are used with a fair train/test protocol.
    Section 4.1 states the datasets differ in degradation, alignment, and resolution; the SOTA claim is limited to these benchmarks.
  • domain assumption Pretrained RAFT-small optical flow, unfrozen during training, provides enough alignment between the LR center image and the reference for warping to help.
    Section 3.2, Eq. (1) warps the reference feature using RAFT flow; poor flow would poison later matching.
  • domain assumption VGG-19 features are a reliable similarity measure for Q-K patch matching across the wide/tele domain gap.
    Section 3.2 uses VGG-19 hidden layers as matching features; the cosine similarity ranking inherits VGG's notion of patch similarity.
  • domain assumption At inference, the LR-center/reference embedding z can substitute for the ground-truth embedding z_gt because the L1 domain-aware loss aligns them.
    Section 3.4 and Eq. (9); no test-time divergence statistics are reported.
  • domain assumption The compared baselines were retrained with a fair protocol and reasonable hyperparameters on the same training set.
    Section 4.1 states 'we retrain all the aforementioned approaches using the same training set as our method'; the fairness of this retraining is not independently audited.
invented entities (2)
  • Domain-aware embedding z (1x1024)
    purpose: Compressed global representation of the LR-center/reference pair used to modulate reconstruction at inference.
    Support is internal: ablation in Table 6 and loss curve in Fig. 5; there is no external falsifiable prediction linked to z.
  • Ground-truth domain embedding z_gt (1x1024)
    purpose: Training-only global representation of the LR/GT pair that supervises z via L1 loss.
    Defined only with access to the ground-truth image, so it cannot be checked outside the training set; its role is to transfer supervision to z.

how reviews work

0 comments
Cite this review

Pith. "Pith review of DM$^3$Net: Dual-Camera Super-Resolution via Domain Modulation and Multi-scale Matching." pith.science (2026). https://pith.science/paper/EUCBFLEK

@misc{pith2026250606993,
  author       = {Pith},
  title        = {Pith review of: DM$^3$Net: Dual-Camera Super-Resolution via Domain Modulation and Multi-scale Matching},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/EUCBFLEK}},
  note         = {Machine review of arXiv:2506.06993}
}
abstract

Dual-camera super-resolution is highly practical for smartphone photography that primarily super-resolve the wide-angle images using the telephoto image as a reference. In this paper, we propose DM$^3$Net, a novel dual-camera super-resolution network based on Domain Modulation and Multi-scale Matching. To bridge the domain gap between the high-resolution domain and the degraded domain, we learn two compressed global representations from image pairs corresponding to the two domains. To enable reliable transfer of high-frequency structural details from the reference image, we design a multi-scale matching module that conducts patch-level feature matching and retrieval across multiple receptive fields to improve matching accuracy and robustness. Moreover, we also introduce Key Pruning to achieve a significant reduction in memory usage and inference time with little model performance sacrificed. Experimental results on three real-world datasets demonstrate that our DM$^3$Net outperforms the state-of-the-art approaches.

Figures

Figures reproduced from arXiv: 2506.06993 by the authors.

Figure 1
Figure 1. Modern multi-lens devices capture scenes at multiple focal lengths simultaneously. By [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Schematics and detailed architectures of DM [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. The flow of (a) the Matching Mechanism and (b) the Key Pruning. [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (8 more)
Figure 4
Figure 4. Figure 4: Reconstructed images produced by the single-scale matching approach, KeDuSR[ [PITH_FULL_IMAGE:figures/full_fig_p005_4.png]
Figure 5
Figure 5. Figure 5: Visualization of the Charbonnier loss [29] curves during training under four experiment settings— (1) without z or zgt, (2) using only z, (3) using only zgt, and (4) using both z and zgt. Motivated by previous image restoration approaches [30, 31], we also learn a comp…
Figure 6
Figure 6. Figure 6: Qualitative comparison results on 3 datasets. Zoom in for best comparison. [PITH_FULL_IMAGE:figures/full_fig_p008_6.png]
Figure 7
Figure 7. Figure 7: PSNR and inference time in relation to varying thresholds under different sampling intervals. [PITH_FULL_IMAGE:figures/full_fig_p014_7.png]
Figure 8
Figure 8. Figure 8: Visual comparisons of ablation study on multi-scale matching. 1/4, 1/2, and 1 denote 1/4 1/2 1 1/4 + 1/2 + 1 GT [PITH_FULL_IMAGE:figures/full_fig_p015_8.png]
Figure 9
Figure 9. Figure 9: Visual comparisons of ablation study on domain modulation. [PITH_FULL_IMAGE:figures/full_fig_p015_9.png]
Figure 10
Figure 10. Figure 10: Visual comparisons on CameraFusion-Real. [PITH_FULL_IMAGE:figures/full_fig_p016_10.png]
Figure 11
Figure 11. Figure 11: Visual comparisons on RealMCVSR-Real [PITH_FULL_IMAGE:figures/full_fig_p017_11.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

40 extracted references · 37 canonical work pages

  1. [1]

    Dual-camera super- resolution with aligned attention modules

    Tengfei Wang, Jiaxin Xie, Wenxiu Sun, Qiong Yan, and Qifeng Chen. Dual-camera super- resolution with aligned attention modules. InProceedings of the IEEE/CVF International Conference on Computer Vision, pages 2001–2010, 2021. 2, 3, 4, 7, 8, 13

    work page 2001

  2. [2]

    Kedusr: real-world dual-lens super- resolution via kernel-free matching

    Huanjing Yue, Zifan Cui, Kun Li, and Jingyu Yang. Kedusr: real-world dual-lens super- resolution via kernel-free matching. InProceedings of the AAAI Conference on Artificial Intelligence, volume 38, pages 6881–6889, 2024. 2, 4, 5, 7, 8, 13

    work page 2024

  3. [3]

    Real-ESRGAN: Training real-world blind super-resolution with pure synthetic data

    Xintao Wang, Liangbin Xie, Chao Dong, and Ying Shan. Real-ESRGAN: Training real-world blind super-resolution with pure synthetic data. InInternational Conference on Computer Vision Workshops. 2, 7, 8

    work page

  4. [4]

    Learning texture trans- former network for image super-resolution

    Fuzhi Yang, Huan Yang, Jianlong Fu, Hongtao Lu, and Baining Guo. Learning texture trans- former network for image super-resolution. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 5791–5800, 2020. 2, 3, 7, 8

    work page 2020

  5. [5]

    CrossNet: An end-to-end reference-based super-resolution network using cross-scale warping

    Haitian Zheng, Mengqi Ji, Haoqian Wang, Yebin Liu, and Lu Fang. CrossNet: An end-to-end reference-based super-resolution network using cross-scale warping. InProceedings of the European Conference on Computer Vision, pages 88–104, 2018

    work page 2018

  6. [6]

    Image super-resolution by neural texture transfer

    Zhifei Zhang, Zhaowen Wang, Zhe Lin, and Hairong Qi. Image super-resolution by neural texture transfer. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2019. 3

    work page 2019

  7. [7]

    Efficient and degradation-adaptive network for real-world image super-resolution

    Jie Liang, Hui Zeng, and Lei Zhang. Efficient and degradation-adaptive network for real-world image super-resolution. InProceedings of the European Conference on Computer Vision, 2022

    work page 2022

  8. [8]

    Robust reference-based super-resolution with similarity-aware deformable convolution

    Gyumin Shim, Jinsun Park, and In So Kweon. Robust reference-based super-resolution with similarity-aware deformable convolution. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, June 2020. 3

    work page 2020

Show all 40 references
  1. [9]

    Landmark image super-resolution by retrieving web images.IEEE Transactions on Image Processing, 22(12):4865–4878, 2013

    Huanjing Yue, Xiaoyan Sun, Jingyu Yang, and Feng Wu. Landmark image super-resolution by retrieving web images.IEEE Transactions on Image Processing, 22(12):4865–4878, 2013. 3

    2013

  2. [10]

    Deformable convolutional networks

    Jifeng Dai, Haozhi Qi, Yuwen Xiong, Yi Li, Guodong Zhang, Han Hu, and Yichen Wei. Deformable convolutional networks. InProceedings of the IEEE international conference on computer vision, pages 764–773, 2017. 3

    2017

  3. [11]

    Deformable convnets v2: More deformable, better results

    Xizhou Zhu, Han Hu, Stephen Lin, and Jifeng Dai. Deformable convnets v2: More deformable, better results. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 9308–9316, 2019. 3

    2019

  4. [12]

    Chan, Xintao Wang, Chen Change Loy, and Ziwei Liu

    Yuming Jiang, Kelvin C.K. Chan, Xintao Wang, Chen Change Loy, and Ziwei Liu. Robust reference-based super-resolution via cˆ2-matching. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2021. 3

    2021

  5. [13]

    Reference-based image super-resolution with deformable attention transformer

    Jiezhang Cao, Jingyun Liang, Kai Zhang, Yawei Li, Yulun Zhang, Wenguan Wang, and Luc Van Gool. Reference-based image super-resolution with deformable attention transformer. In Proceedings of the European Conference on Computer Vision, 2022. 3

    2022

  6. [14]

    Masa-sr: Matching acceleration and spatial adaptation for reference-based image super-resolution

    Liying Lu, Wenbo Li, Xin Tao, Jiangbo Lu, and Jiaya Jia. Masa-sr: Matching acceleration and spatial adaptation for reference-based image super-resolution. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2021. 3, 7, 8

    2021

  7. [15]

    Gomez, Łukasz Kaiser, and Illia Polosukhin

    Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N. Gomez, Łukasz Kaiser, and Illia Polosukhin. Attention is all you need. InAdvances in Neural Informa- tion Processing Systems, pages 5998–6008, 2017. 3

    2017

  8. [16]

    Coarse- to-fine embedded patchmatch and multi-scale dynamic aggregation for reference-based super- resolution

    Bin Xia, Yapeng Tian, Yucheng Hang, Wenming Yang, Qingmin Liao, and Jie Zhou. Coarse- to-fine embedded patchmatch and multi-scale dynamic aggregation for reference-based super- resolution. InProceedings of the AAAI Conference on Artificial Intelligence, volume 36, pages 2768–2...

    2022

  9. [17]

    Image super-resolution using deep convolutional networks.IEEE Transactions on Pattern Analysis and Machine Intelligence, 38(2):295–307, 2016

    Chao Dong, Chen Change Loy, Kaiming He, and Xiaoou Tang. Image super-resolution using deep convolutional networks.IEEE Transactions on Pattern Analysis and Machine Intelligence, 38(2):295–307, 2016. 3

    2016

  10. [18]

    Brightness and color correction for dual camera image registration

    Joo Park, Sung Kim, and Hyun Kang. Brightness and color correction for dual camera image registration. InProceedings of IEEE International Conference on Consumer Electronics–Asia (ICCE-Asia), pages 1–4, 2016. 3

    2016

  11. [19]

    ORB: An efficient alternative to SIFT or SURF

    Ethan Rublee, Vincent Rabaud, Kurt Konolige, and Gary Bradski. ORB: An efficient alternative to SIFT or SURF. InProceedings of IEEE International Conference on Computer Vision, pages 2564–2571, 2011

    2011

  12. [20]

    Low-light color imaging via dual camera acquisition

    Xiaojie Guo, Yifan Li, Zhe Yu, and Jiaya Liu. Low-light color imaging via dual camera acquisition. InAsian Conference on Computer Vision (ACCV), pages 123–138, 2020. 3

    2020

  13. [21]

    Self-supervised learning for real-world super-resolution from dual zoomed observations

    Zhilu Zhang, Ruohao Wang, Hongzhi Zhang, Yunjin Chen, and Wangmeng Zuo. Self-supervised learning for real-world super-resolution from dual zoomed observations. InProceedings of the European Conference on Computer Vision, 2022. 3, 7, 8, 13

    2022

  14. [22]

    Zeroshot duallens superresolution

    Ruikang Xu, Mingde Yao, and Zhiwei Xiong. Zeroshot duallens superresolution. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2023. 3

    2023

  15. [23]

    Efficient hybrid zoom using camera fusion on mobile phones.ACM Trans

    Xiaotong Wu, Wei-Sheng Lai, Yichang Shih, Charles Herrmann, Michael Krainin, Deqing Sun, and Chia-Kai Liang. Efficient hybrid zoom using camera fusion on mobile phones.ACM Trans. Graph., 42(4):1–15, 2023. 4

    2023

  16. [24]

    Deep residual learning for image recognition

    Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. Deep residual learning for image recognition. InProceedings of the IEEE conference on computer vision and pattern recognition, pages 770–778, 2016. 4

    2016

  17. [25]

    Squeeze-and-excitation networks

    Jie Hu, Li Shen, and Gang Sun. Squeeze-and-excitation networks. In2018 IEEE/CVF Confer- ence on Computer Vision and Pattern Recognition, pages 7132–7141, 2018. 4

    2018

  18. [26]

    Raft: Recurrent all-pairs field transforms for optical flow

    Zachary Teed and Jia Deng. Raft: Recurrent all-pairs field transforms for optical flow. In Proceedings of the European Conference on Computer Vision, 2020. 4

    2020

  19. [27]

    Very deep convolutional networks for large-scale image recognition

    Karen Simonyan and Andrew Zisserman. Very deep convolutional networks for large-scale image recognition. InInternational Conference on Learning Representations (ICLR), 2015. 4, 13

    2015

  20. [28]

    Deep learning using rectified linear units (relu).arXiv preprint arXiv:1803.08375, 2018

    Abien Fred Agarap. Deep learning using rectified linear units (relu).arXiv preprint arXiv:1803.08375, 2018. 5

    2018 arXiv

  21. [29]

    Charbonnier et al

    P. Charbonnier et al. Two-dimensional signal and image processing.IEEE Transactions on Image Processing, 3(2):68–72, 1994. 6, 13

    1994

  22. [30]

    Diffir: Efficient diffusion model for image restoration.Proceedings of IEEE International Conference on Computer Vision, 2023

    Bin Xia, Yulun Zhang, Shiyin Wang, Yitong Wang, Xinglong Wu, Yapeng Tian, Wenming Yang, and Luc Van Gool. Diffir: Efficient diffusion model for image restoration.Proceedings of IEEE International Conference on Computer Vision, 2023. 6

    2023

  23. [31]

    Generating content for hdr deghost- ing from frequency view

    Tao Hu, Qingsen Yan, Yuankai Qi, and Yanning Zhang. Generating content for hdr deghost- ing from frequency view. In2024 IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 25732–25741, 2024. 6

    2024

  24. [32]

    Prentice Hall PTR, 1994

    Simon Haykin.Neural networks: a comprehensive foundation. Prentice Hall PTR, 1994. 6

    1994

  25. [33]

    Arbitrary style transfer in real-time with adaptive instance normalization

    Xun Huang and Serge Belongie. Arbitrary style transfer in real-time with adaptive instance normalization. InProceedings of the IEEE International Conference on Computer Vision, Oct

  26. [34]

    Adam: A method for stochastic optimization.arXiv preprint arXiv:1412.6980, 2014

    Diederik P Kingma and Jimmy Ba. Adam: A method for stochastic optimization.arXiv preprint arXiv:1412.6980, 2014. 7 11

    2014 arXiv

  27. [35]

    Pytorch: An imperative style, high-performance deep learning library

    Adam Paszke, Sam Gross, Francisco Massa, Adam Lerer, James Bradbury, Gregory Chanan, Trevor Killeen, Zeming Lin, Natalia Gimelshein, Luca Antiga, Alban Desmaison, Andreas Kopf, Edward Yang, Zachary DeVito, Martin Raison, Alykhan Tejani, Sasank Chilamkurthy, Benoit Steiner, Lu ...

    2019

  28. [36]

    Hore and D

    A. Hore and D. Ziou. Image quality metrics: Psnr vs. ssim—putting structural similarity on a sound mathematical foundation. InProceedings of the 2010 International Conference on Pattern Recognition (ICPR), pages 2366–2369, 2010. 7

    2010

  29. [37]

    Z. Wang, A. C. Bovik, H. R. Sheikh, and E. P. Simoncelli. Image quality assessment: From error visibility to structural similarity.IEEE Transactions on Image Processing, 13(4):600–612,

  30. [38]

    Efros, Eli Shechtman, and Oliver Wang

    Richard Zhang, Phillip Isola, Alexei A. Efros, Eli Shechtman, and Oliver Wang. The unrea- sonable effectiveness of deep features as a perceptual metric. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 586–595, 2018. 7

    2018

  31. [39]

    Swinir: Image restoration using swin transformer.arXiv preprint arXiv:2108.10257, 2021

    Jingyun Liang, Jiezhang Cao, Guolei Sun, Kai Zhang, Luc Van Gool, and Radu Timofte. Swinir: Image restoration using swin transformer.arXiv preprint arXiv:2108.10257, 2021. 7, 8

    2021 arXiv

  32. [40]

    Perceptual losses for real-time style transfer and super-resolution.Proceedings of the European Conference on Computer Vision, pages 694–711, 2016

    Justin Johnson, Alexandre Alahi, and Li Fei-Fei. Perceptual losses for real-time style transfer and super-resolution.Proceedings of the European Conference on Computer Vision, pages 694–711, 2016. 13 12 Appendix and supplemental material A Loss Functions In the training phase,...

    2016

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.