pith. machine review for the scientific record. sign in

arxiv: 2605.02184 · v1 · submitted 2026-05-04 · 💻 cs.CV · cs.AI· cs.LG

Recognition: unknown

RAFNet: Region-Aware Fusion Network for Pansharpening

Jianing Zhang , Zijian Zhou , Kai Sun
Authors on Pith no claims yet

Pith reviewed 2026-05-09 16:33 UTC · model grok-4.3

classification 💻 cs.CV cs.AIcs.LG
keywords pansharpeningimage fusiondeep learningregion-aware processingwavelet transformsparse attentionremote sensing imagery
0
0 comments X

The pith

Region-specific adaptive kernels and sparse attention let a fusion network outperform prior pansharpening methods on standard benchmarks.

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

The paper sets out to show that partitioning remote-sensing images into semantic regions and tailoring both convolution kernels and attention operations to those regions produces higher-quality fused multispectral images while lowering computational cost. Existing frequency methods incur quadratic complexity and ignore regional differences; static spatial kernels cannot adjust to varying frequency content across a scene. By using wavelet decomposition to isolate directional frequencies, clustering to define regions, and then building per-region adaptive kernels plus cluster-guided sparse attention, the network is claimed to capture the necessary spatial-frequency interactions more efficiently. A reader would care because accurate pansharpening directly improves the utility of satellite imagery for mapping and monitoring.

Core claim

RAFNet integrates a Spatial Adaptive Refinement module that applies discrete wavelet transform for frequency separation followed by K-means clustering to partition the image and then constructs region-specific adaptive convolution kernels, together with a Clustered Frequency Aggregation module that performs semantic-cluster-guided sparse attention to aggregate frequency features with reduced redundancy; these two modules are embedded in a progressive multi-level spatial-frequency architecture that enables repeated interaction between the adapted features, yielding measurably better reconstruction of high-resolution multispectral images.

What carries the argument

The pair of SAR and CFA modules that together produce region-partitioned, frequency-aware adaptive convolution and sparse attention.

If this is right

  • Pansharpened outputs improve on both reduced-resolution and full-resolution benchmark protocols.
  • Sparse attention guided by semantic clusters cuts quadratic complexity while preserving frequency detail.
  • Dynamic kernels per region adapt to the directional frequency content of PAN and MS imagery.
  • Progressive multi-level fusion allows repeated refinement between spatial and frequency streams.

Where Pith is reading between the lines

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

  • Similar region-guided adaptation could be tested on other fusion problems such as hyperspectral sharpening or medical image registration.
  • The clustering step introduces a discrete partitioning that might be replaced by differentiable soft assignment to allow end-to-end gradient flow.
  • If the method generalizes, it could reduce the need for heavy post-processing in operational remote-sensing pipelines.

Load-bearing premise

The reported gains come from the region-adaptive SAR and CFA modules themselves rather than from differences in training schedules, data augmentation, or metric selection.

What would settle it

Re-training the strongest competing methods under identical data splits, augmentation, and optimization settings and finding that the performance gap disappears would falsify the claim that the new modules are responsible for the improvement.

Figures

Figures reproduced from arXiv: 2605.02184 by Jianing Zhang, Kai Sun, Zijian Zhou.

Figure 1
Figure 1. Figure 1: Evolution of convolution paradigms for spatial feature enhancement. (a) Static Conv applies a spatially invariant filter across the entire image, often view at source ↗
Figure 2
Figure 2. Figure 2: Motivation for the proposed Cluster-Routed Sparse Attention (CRSA). (a) CDF Energy Curve: Demonstrates the long-tail distribution of attention view at source ↗
Figure 3
Figure 3. Figure 3: Overall architecture of the proposed Region-Aware Fusion (RAF) network. The network takes PAN and LRMS images as inputs and primarily consists view at source ↗
Figure 4
Figure 4. Figure 4: Detailed architectures the Spatial Adaptive Convolution (SAC) unit within the SAR module. It comprises the Cluster Region Partition for generating view at source ↗
Figure 5
Figure 5. Figure 5: Architectures of the Clustered Frequency Aggregation (CFA) module, detailing the Cluster-Routed Sparse Attention (CRSA) and the GDFN. The view at source ↗
Figure 6
Figure 6. Figure 6: The visual results (Top) and residuals (Bottom) of all compared approaches on the WorldView-3 reduced-resolution dataset. view at source ↗
Figure 7
Figure 7. Figure 7: The visual results (Top) and residuals (Bottom) of all compared approaches on the Gaofen-2 reduced-resolution dataset. view at source ↗
Figure 8
Figure 8. Figure 8: The visual results of all compared approaches on the GaoFen-2 full-resolution dataset. view at source ↗
Figure 9
Figure 9. Figure 9: Visual representations of cluster index matrices in the model at different training epochs. (a) and (b) show the visual appearance of the raw PAN and view at source ↗
Figure 10
Figure 10. Figure 10: Variations of PSNR, SAM and ERGAS on the WorldView-3 reduced-resolution dataset with changing cluster number view at source ↗
read the original abstract

Pansharpening aims to generate high-resolution multispectral (HRMS) images by fusing low-resolution multispectral (LRMS) and high-resolution panchromatic (PAN) images. Although deep learning has advanced this field, mainstream frequency-based methods relying on standard scaled dot-product attention suffer from quadratic computational complexity and fail to exploit the inherent regional sparsity of remote sensing imagery. Furthermore, existing spatial enhancement strategies typically employ static convolution kernels, which struggle to adapt to the complex frequency and regional variations of PAN and MS images. To address these bottlenecks, we propose a Region-Aware Fusion (RAFNet) Network that synergistically models spatial and frequency information. Specifically, we design a Spatial Adaptive Refinement (SAR) module that leverages the discrete wavelet transform (DWT) for directional frequency separation and K-means clustering for regional partitioning, which enables the dynamic construction of region-specific adaptive convolution kernels, achieving spatially and frequency-adaptive feature enhancement. Moreover, we introduce a Clustered Frequency Aggregation (CFA) module based on a sparse attention mechanism guided by the semantic clusters, which executes a region-aware sparse attention strategy that drastically reduces computational redundancy while ensuring high-quality frequency feature extraction. In addition we integrated these modules into a progressive, multi-level spatial-frequency network architecture to facilitate robust interaction and accurate image reconstruction. Extensive experiments on multiple benchmark datasets demonstrate that the proposed RAFNet significantly outperforms state-of-the-art pansharpening methods in both reduced- and full-resolution assessments. The code is available at https://github.com/PatrickNod/RAFNet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The paper proposes RAFNet for pansharpening, fusing low-resolution multispectral (LRMS) and high-resolution panchromatic (PAN) images via a multi-level spatial-frequency architecture. It introduces a Spatial Adaptive Refinement (SAR) module that applies discrete wavelet transform (DWT) for frequency separation and K-means for regional partitioning to build dynamic, region-specific adaptive convolution kernels, plus a Clustered Frequency Aggregation (CFA) module that uses cluster-guided sparse attention to reduce quadratic complexity while preserving frequency features. Extensive experiments on benchmark datasets are claimed to show significant outperformance over state-of-the-art methods in both reduced- and full-resolution settings, with code released.

Significance. If the reported gains are shown to arise specifically from the SAR and CFA modules under controlled conditions, the work would usefully address two persistent limitations in pansharpening—static kernels that ignore regional frequency variation and the O(n²) cost of standard attention—potentially improving both quality and efficiency for remote-sensing applications.

major comments (2)
  1. [§4] §4 (Experiments) and associated tables: the central claim that RAFNet 'significantly outperforms' SOTA methods requires explicit confirmation that all baselines were retrained from scratch under identical protocols (optimizer, learning-rate schedule, epoch count, data augmentation, loss weighting, and post-processing). If the comparisons rely on published numbers without matched retraining, the attribution of gains to the SAR (DWT+K-means) and CFA (cluster-guided sparse attention) modules remains unisolated and the empirical evidence is inconclusive.
  2. [§3.2] §3.2 (SAR module description): the claim that K-means clustering enables 'dynamic construction of region-specific adaptive convolution kernels' is load-bearing for the spatial-adaptive contribution, yet the manuscript provides no ablation that isolates the effect of the clustering step versus the DWT frequency separation alone, nor any analysis of sensitivity to the choice of K or the clustering feature space.
minor comments (2)
  1. [Abstract] Abstract: the sentence 'In addition we integrated these modules...' is grammatically incomplete and should be revised for clarity.
  2. [§3] The manuscript would benefit from a concise table summarizing the computational complexity (FLOPs and parameters) of RAFNet versus the compared attention-based baselines to quantify the claimed reduction in redundancy.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive and detailed comments on our manuscript. We address each major comment below and commit to revisions that strengthen the empirical validation and module analysis.

read point-by-point responses

  1. Referee: [§4] §4 (Experiments) and associated tables: the central claim that RAFNet 'significantly outperforms' SOTA methods requires explicit confirmation that all baselines were retrained from scratch under identical protocols (optimizer, learning-rate schedule, epoch count, data augmentation, loss weighting, and post-processing). If the comparisons rely on published numbers without matched retraining, the attribution of gains to the SAR (DWT+K-means) and CFA (cluster-guided sparse attention) modules remains unisolated and the empirical evidence is inconclusive.

    Authors: We agree that matched retraining under identical protocols is essential for isolating the contributions of SAR and CFA. All baselines were retrained from scratch using the same Adam optimizer, cosine-annealing learning-rate schedule, 200 epochs, identical data augmentations (random horizontal/vertical flips and rotations), the same combined loss with equal weighting of terms, and no post-processing. To make this fully transparent, we will add a new subsection in the revised §4 that explicitly lists the common training configuration and confirms each baseline was optimized under these conditions. Updated tables will reference these protocols. revision: yes

  2. Referee: [§3.2] §3.2 (SAR module description): the claim that K-means clustering enables 'dynamic construction of region-specific adaptive convolution kernels' is load-bearing for the spatial-adaptive contribution, yet the manuscript provides no ablation that isolates the effect of the clustering step versus the DWT frequency separation alone, nor any analysis of sensitivity to the choice of K or the clustering feature space.

    Authors: We acknowledge that an explicit ablation isolating the K-means step from DWT alone, together with sensitivity analysis, would strengthen the justification for the clustering component. We have run additional controlled experiments: (i) SAR with DWT only versus full SAR (DWT + K-means), and (ii) sensitivity sweeps over K ∈ {2,4,8,16} and alternative clustering feature spaces (spatial statistics versus wavelet coefficients). These results show measurable gains from clustering in heterogeneous regions and will be incorporated into the revised manuscript as a new table and accompanying analysis in §4. revision: yes

Circularity Check

0 steps flagged

No circularity: empirical claims rest on benchmark comparisons without self-referential reductions

full rationale

The paper introduces RAFNet as a novel architecture combining SAR (DWT + K-means for adaptive kernels) and CFA (cluster-guided sparse attention) modules within a multi-level spatial-frequency network. All load-bearing assertions concern empirical outperformance on reduced- and full-resolution pansharpening benchmarks, with no equations, first-principles derivations, or predictions that reduce by construction to fitted parameters, self-definitions, or self-citation chains. The design choices are presented as motivated engineering decisions validated externally via code release and dataset comparisons, rendering the work self-contained against independent benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

The central claim rests on the empirical superiority of the SAR and CFA modules. No explicit free parameters beyond standard network weights are named in the abstract. No new physical axioms or invented entities are introduced.

pith-pipeline@v0.9.0 · 5582 in / 1162 out tokens · 42657 ms · 2026-05-09T16:33:47.392675+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

56 extracted references · 6 canonical work pages · 4 internal anchors

  1. [1]

    Field-derived spectra of salinized soils and vegetation as indicators of irrigation-induced soil salinization,

    R. Dehaan and G. Taylor, “Field-derived spectra of salinized soils and vegetation as indicators of irrigation-induced soil salinization,”Remote Sens. Environ., vol. 80, no. 3, pp. 406–417, 2002

    2002

  2. [2]

    Application of hyperspectral remote sensing for environment monitoring in mining areas,

    B. Zhang, D. Wu, L. Zhang, Q. Jiao, and Q. Li, “Application of hyperspectral remote sensing for environment monitoring in mining areas,”Environ. Earth Sci., vol. 65, no. 3, pp. 649–658, 2012

    2012

  3. [3]

    Review of the pansharpening methods for remote sensing images based on the idea of meta-analysis: Practical discussion and challenges,

    X. Meng, H. Shen, H. Li, L. Zhang, and R. Fu, “Review of the pansharpening methods for remote sensing images based on the idea of meta-analysis: Practical discussion and challenges,”Inf. Fusion, vol. 46, pp. 102–113, 2019

    2019

  4. [4]

    A new adaptive component-substitution- based satellite image fusion by using partial replacement,

    J. Choi, K. Yu, and Y . Kim, “A new adaptive component-substitution- based satellite image fusion by using partial replacement,”IEEE Trans. Geosci. Remote Sens., vol. 49, no. 1, pp. 295–309, 2010

    2010

  5. [5]

    Robust band-dependent spatial-detail approaches for panchromatic sharpening,

    G. Vivone, “Robust band-dependent spatial-detail approaches for panchromatic sharpening,”IEEE Trans. Geosci. Remote Sens., vol. 57, no. 9, pp. 6421–6433, 2019

    2019

  6. [6]

    Contrast and error-based fusion schemes for multispectral image pansharpening,

    G. Vivone, R. Restaino, M. Dalla Mura, G. Licciardi, and J. Chanus- sot, “Contrast and error-based fusion schemes for multispectral image pansharpening,”IEEE Trans. Geosci. Remote Sens., vol. 11, no. 5, pp. 930–934, 2013

    2013

  7. [7]

    Full scale regression-based injection coefficients for panchromatic sharpening,

    G. Vivone, R. Restaino, and J. Chanussot, “Full scale regression-based injection coefficients for panchromatic sharpening,”IEEE Trans. Image Process., vol. 27, no. 7, pp. 3418–3431, 2018

    2018

  8. [8]

    A variational pan-sharpening with local gradient constraints,

    X. Fu, Z. Lin, Y . Huang, and X. Ding, “A variational pan-sharpening with local gradient constraints,” inProc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit. (CVPR), 2019, pp. 10 265–10 274

    2019

  9. [9]

    Variational pansharpening by exploiting cartoon-texture similarities,

    X. Tian, Y . Chen, C. Yang, and J. Ma, “Variational pansharpening by exploiting cartoon-texture similarities,”IEEE Trans. Geosci. Remote Sens., vol. 60, pp. 1–16, 2021

    2021

  10. [10]

    A critical comparison among IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. XX, NO. XX, [MONTH] 2026 12 pansharpening algorithms,

    G. Vivone, L. Alparone, J. Chanussot, M. Dalla Mura, A. Garzelli, G. A. Licciardi, R. Restaino, and L. Wald, “A critical comparison among IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. XX, NO. XX, [MONTH] 2026 12 pansharpening algorithms,”IEEE Trans. Geosci. Remote Sens., vol. 53, no. 5, pp. 2565–2586, 2014

    2026

  11. [11]

    Wavelet based image fusion techniques—an introduction, review and comparison,

    K. Amolins, Y . Zhang, and P. Dare, “Wavelet based image fusion techniques—an introduction, review and comparison,”ISPRS J. Pho- togramm. Remote Sens., vol. 62, no. 4, pp. 249–263, 2007

    2007

  12. [12]

    A theoretical and practical survey of image fusion methods for multispectral pansharpening,

    C. S. Yilmaz, V . Yilmaz, and O. Gungor, “A theoretical and practical survey of image fusion methods for multispectral pansharpening,”Inf. Fusion, vol. 79, pp. 1–43, 2022

    2022

  13. [13]

    Deep unsupervised blind hyperspectral and multispectral data fusion,

    J. Li, K. Zheng, J. Yao, L. Gao, and D. Hong, “Deep unsupervised blind hyperspectral and multispectral data fusion,”IEEE Geosci. Remote Sens. Lett., vol. 19, pp. 1–5, 2022

    2022

  14. [14]

    X-shaped interactive autoencoders with cross-modality mutual learning for unsupervised hyperspectral image super-resolution,

    J. Li, K. Zheng, Z. Li, L. Gao, and X. Jia, “X-shaped interactive autoencoders with cross-modality mutual learning for unsupervised hyperspectral image super-resolution,”IEEE Trans. Geosci. Remote Sens., vol. 61, pp. 1–17, 2023

    2023

  15. [15]

    Pansharpening by convolutional neural networks,

    G. Masi, D. Cozzolino, L. Verdoliva, and G. Scarpa, “Pansharpening by convolutional neural networks,”Remote Sens., vol. 8, no. 7, p. 594, 2016

    2016

  16. [16]

    Pan- sharpening via detail injection based convolutional neural networks,

    L. He, Y . Rao, J. Li, J. Chanussot, A. Plaza, J. Zhu, and B. Li, “Pan- sharpening via detail injection based convolutional neural networks,” IEEE J. Sel. Topics Appl. Earth Observ. Remote Sens., vol. 12, no. 4, pp. 1188–1204, 2019

    2019

  17. [17]

    Detail injection-based deep convolutional neural networks for pansharpening,

    L.-J. Deng, G. Vivone, C. Jin, and J. Chanussot, “Detail injection-based deep convolutional neural networks for pansharpening,”IEEE Trans. Geosci. Remote Sens., vol. 59, no. 8, pp. 6995–7010, 2020

    2020

  18. [18]

    U2net: A general framework with spatial-spectral-integrated double u-net for image fusion,

    S. Peng, C. Guo, X. Wu, and L.-J. Deng, “U2net: A general framework with spatial-spectral-integrated double u-net for image fusion,” inProc. ACM Int. Conf. Multimedia (MM), 2023, pp. 3219–3227

    2023

  19. [19]

    Pixel-adaptive convolutional neural networks,

    H. Su, V . Jampani, D. Sun, O. Gallo, E. Learned-Miller, and J. Kautz, “Pixel-adaptive convolutional neural networks,” inProc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit. (CVPR), 2019, pp. 11 166–11 175

    2019

  20. [20]

    On the performance of convolutional neural networks under high and low frequency information,

    R. R. Yedla and S. R. Dubey, “On the performance of convolutional neural networks under high and low frequency information,” inProc. Int. Conf. Comput. Vis. Image Process. (CVIP), 2020, pp. 214–224

    2020

  21. [21]

    Frequency separation for real-world super-resolution,

    M. Fritsche, S. Gu, and R. Timofte, “Frequency separation for real-world super-resolution,” inProc. IEEE/CVF Int. Conf. Comput. Vis. Workshop (ICCVW), 2019, pp. 3599–3608

    2019

  22. [22]

    Wavelet- assisted multi-frequency attention network for pansharpening,

    J. Huang, R. Huang, J. Xu, S. Peng, Y . Duan, and L.-J. Deng, “Wavelet- assisted multi-frequency attention network for pansharpening,” inProc. AAAI Conf. Artif. Intell., vol. 39, no. 4, 2025, pp. 3662–3670

    2025

  23. [23]

    Efficient language modeling with sparse all-mlp,

    Z. Zhou, Y . Tay, R. Nallapati, B. Mitra, Z. Xiao, H. Cheng, X. Xiang, J. P. Sim, H. Swaminathan, N. D. Tranet al., “Efficient language modeling with sparse all-mlp,”arXiv preprint arXiv:2203.06850, 2022

    work page arXiv 2022

  24. [24]

    Dynamic filter networks,

    X. Jia, B. De Brabandere, T. Tuytelaars, and L. V . Gool, “Dynamic filter networks,” inProc. Adv. Neural Inf. Process. Syst. (NIPS), vol. 29, 2016

    2016

  25. [25]

    Decoupled dynamic filter networks,

    J. Zhou, V . Jampani, Z. Pi, Q. Liu, and M.-H. Yang, “Decoupled dynamic filter networks,” inProc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit. (CVPR), 2021, pp. 6647–6656

    2021

  26. [26]

    Cross-scale internal graph neural network for image super-resolution,

    S. Zhou, J. Zhang, W. Zuo, and C. C. Loy, “Cross-scale internal graph neural network for image super-resolution,” inProc. Adv. Neural Inf. Process. Syst. (NIPS), vol. 33, 2020, pp. 3499–3509

    2020

  27. [27]

    Content-adaptive non- local convolution for remote sensing pansharpening,

    Y . Duan, X. Wu, H. Deng, and L.-J. Deng, “Content-adaptive non- local convolution for remote sensing pansharpening,” inProc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit. (CVPR), 2024, pp. 27 738–27 747

    2024

  28. [28]

    Frequency dynamic convolution for dense image prediction,

    L. Chen, L. Gu, L. Li, C. Yan, and Y . Fu, “Frequency dynamic convolution for dense image prediction,” inProc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit. (CVPR), 2025, pp. 30 178–30 188

    2025

  29. [29]

    A general paradigm with detail-preserving conditional invertible network for image fusion,

    W. Wang, L.-J. Deng, R. Ran, and G. Vivone, “A general paradigm with detail-preserving conditional invertible network for image fusion,”Int. J. Comput. Vis., vol. 132, no. 4, pp. 1029–1054, 2024

    2024

  30. [30]

    Fsgformer: Frequency separation and guidance transformer for pansharpening,

    Q. Liu, X. Zhao, Y . Qin, L. Li, and J. Liu, “Fsgformer: Frequency separation and guidance transformer for pansharpening,”IEEE Trans. Geosci. Remote Sens., 2025

    2025

  31. [31]

    Mscscformer: Multiscale convolutional sparse coding-based transformer for pansharpening,

    Y . Ye, T. Wang, F. Fang, and G. Zhang, “Mscscformer: Multiscale convolutional sparse coding-based transformer for pansharpening,”IEEE Trans. Geosci. Remote Sens., vol. 62, pp. 1–12, 2024

    2024

  32. [32]

    Lite transformer with long-short range attention

    Z. Wu, Z. Liu, J. Lin, Y . Lin, and S. Han, “Lite transformer with long- short range attention,”arXiv preprint arXiv:2004.11886, 2020

    work page arXiv 2004

  33. [33]

    Longformer: The Long-Document Transformer

    I. Beltagy, M. E. Peters, and A. Cohan, “Longformer: The long- document transformer,”arXiv preprint arXiv:2004.05150, 2020

    work page internal anchor Pith review arXiv 2004

  34. [34]

    Etc: Encoding long and structured inputs in transformers,

    J. Ainslie, S. Ontanon, C. Alberti, V . Cvicek, Z. Fisher, P. Pham, A. Ravula, S. Sanghai, Q. Wang, and L. Yang, “Etc: Encoding long and structured inputs in transformers,” inProc. Conf. Empir. Methods Nat. Lang. Process. (EMNLP), 2020, pp. 268–284

    2020

  35. [35]

    Representation Learning with Contrastive Predictive Coding

    A. v. d. Oord, Y . Li, and O. Vinyals, “Representation learning with contrastive predictive coding,”arXiv preprint arXiv:1807.03748, 2018

    work page internal anchor Pith review Pith/arXiv arXiv 2018

  36. [36]

    Reformer: The Efficient Transformer

    N. Kitaev, Ł. Kaiser, and A. Levskaya, “Reformer: The efficient trans- former,”arXiv preprint arXiv:2001.04451, 2020

    work page internal anchor Pith review arXiv 2001

  37. [37]

    Efficient content- based sparse attention with routing transformers,

    A. Roy, M. Saffar, A. Vaswani, and D. Grangier, “Efficient content- based sparse attention with routing transformers,”Trans. Assoc. Comput. Linguist., vol. 9, pp. 53–68, 2021

    2021

  38. [38]

    Pan-guided multiresolution fusion network using swin transformer for pansharpening,

    L. Hou, B. Zhang, and B. Wang, “Pan-guided multiresolution fusion network using swin transformer for pansharpening,”IEEE Geosci. Remote Sens. Lett., vol. 20, pp. 1–5, 2022

    2022

  39. [39]

    Dynamic region-aware convolution,

    J. Chen, X. Wang, Z. Guo, X. Zhang, and J. Sun, “Dynamic region-aware convolution,” inProc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit. (CVPR), 2021, pp. 8064–8073

    2021

  40. [40]

    Dynamic convolution: Attention over convolution kernels,

    Y . Chen, X. Dai, M. Liu, D. Chen, L. Yuan, and Z. Liu, “Dynamic convolution: Attention over convolution kernels,” inProc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit. (CVPR), 2020, pp. 11 030–11 039

    2020

  41. [41]

    Lagconv: Local-context adaptive convolution kernels with global harmonic bias for pansharpening,

    Z.-R. Jin, T.-J. Zhang, T.-X. Jiang, G. Vivone, and L.-J. Deng, “Lagconv: Local-context adaptive convolution kernels with global harmonic bias for pansharpening,” inProc. AAAI Conf. Artif. Intell., vol. 36, no. 1, 2022, pp. 1113–1121

    2022

  42. [42]

    Attention is all you need,

    A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A. N. Gomez, Ł. Kaiser, and I. Polosukhin, “Attention is all you need,” inProc. Adv. Neural Inf. Process. Syst. (NIPS), vol. 30, 2017

    2017

  43. [43]

    An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale

    A. Dosovitskiy, L. Beyer, A. Kolesnikov, D. Weissenborn, X. Zhai, T. Unterthiner, M. Dehghani, M. Minderer, G. Heigold, S. Gellyet al., “An image is worth 16x16 words: Transformers for image recognition at scale,”arXiv preprint arXiv:2010.11929, 2020

    work page internal anchor Pith review Pith/arXiv arXiv 2010

  44. [44]

    Vision transformer for pansharpening,

    X. Meng, N. Wang, F. Shao, and S. Li, “Vision transformer for pansharpening,”IEEE Trans. Geosci. Remote Sens., vol. 60, pp. 1–11, 2022

    2022

  45. [45]

    Hypertransformer: A textural and spectral feature fusion transformer for pansharpening,

    W. G. C. Bandara and V . M. Patel, “Hypertransformer: A textural and spectral feature fusion transformer for pansharpening,” inProc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit. (CVPR), 2022, pp. 1767–1777

    2022

  46. [46]

    Restormer: Efficient transformer for high-resolution image restoration,

    S. W. Zamir, A. Arora, S. Khan, M. Hayat, F. S. Khan, and M.-H. Yang, “Restormer: Efficient transformer for high-resolution image restoration,” inProc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit. (CVPR), 2022, pp. 5728–5739

    2022

  47. [47]

    Dkm: Differentiable k-means clustering layer for neural network compression,

    M. Cho, K. A. Vahid, S. Adya, and M. Rastegari, “Dkm: Differentiable k-means clustering layer for neural network compression,” inInt. Conf. Learn. Represent., 2012, pp. 2061–2079

    2012

  48. [48]

    Fusion of satellite images of different spatial resolutions: Assessing the quality of resulting images,

    L. Wald, T. Ranchin, and M. Mangolini, “Fusion of satellite images of different spatial resolutions: Assessing the quality of resulting images,” Photogramm. Eng. Remote Sens., vol. 63, no. 6, pp. 691–699, 1997

    1997

  49. [49]

    Process for enhancing the spatial resolution of multispectral imagery using pan-sharpening,

    C. A. Laben and B. V . Brower, “Process for enhancing the spatial resolution of multispectral imagery using pan-sharpening,” U.S. Patent 6 011 875, Jan. 4, 2000

    2000

  50. [50]

    Optimal mmse pan sharpening of very high resolution multispectral images,

    A. Garzelli, F. Nencini, and L. Capobianco, “Optimal mmse pan sharpening of very high resolution multispectral images,”IEEE Trans. Geosci. Remote Sens., vol. 46, no. 1, pp. 228–236, 2007

    2007

  51. [51]

    Pannet: A deep network architecture for pan-sharpening,

    J. Yang, X. Fu, Y . Hu, Y . Huang, X. Ding, and J. Paisley, “Pannet: A deep network architecture for pan-sharpening,” inProc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit. (CVPR), 2017, pp. 5449–5457

    2017

  52. [52]

    Deep gradient projection networks for pan-sharpening,

    S. Xu, J. Zhang, Z. Zhao, K. Sun, J. Liu, and C. Zhang, “Deep gradient projection networks for pan-sharpening,” inProc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit. (CVPR), 2021, pp. 1366–1375

    2021

  53. [53]

    Frequency- adaptive pan-sharpening with mixture of experts,

    X. He, K. Yan, R. Li, C. Xie, J. Zhang, and M. Zhou, “Frequency- adaptive pan-sharpening with mixture of experts,” inProc. AAAI Conf. Artif. Intell., vol. 38, no. 3, 2024, pp. 2121–2129

    2024

  54. [54]

    Dcpnet: A dual- task collaborative promotion network for pansharpening,

    Y . Zhang, X. Yang, H. Li, M. Xie, and Z. Yu, “Dcpnet: A dual- task collaborative promotion network for pansharpening,”IEEE Trans. Geosci. Remote Sens., vol. 62, pp. 1–16, 2024

    2024

  55. [55]

    Panchromatic and multispectral image fusion for remote sensing and earth observation: Concepts, taxonomy, literature review, evaluation methodologies and challenges ahead,

    K. Zhang, F. Zhang, W. Wan, H. Yu, J. Sun, J. Del Ser, E. Elyan, and A. Hussain, “Panchromatic and multispectral image fusion for remote sensing and earth observation: Concepts, taxonomy, literature review, evaluation methodologies and challenges ahead,”Inf. Fusion, vol. 93, pp. 227–242, 2023

    2023

  56. [56]

    Full-resolution quality assessment of pansharpening: Theoretical and hands-on approaches,

    A. Arienzo, G. Vivone, A. Garzelli, L. Alparone, and J. Chanussot, “Full-resolution quality assessment of pansharpening: Theoretical and hands-on approaches,”IEEE Geosci. Remote Sens. Mag., vol. 10, no. 3, pp. 168–201, 2022

    2022