One persistent challenge in remote sensing is boosting the spatial clarity of openly available Sentinel‑2 imagery to a level on par with commercial products. To meet this challenge, we present KDeHAT‑SR, which integrates degradation kernel estimation with a hierarchical attention Transformer for super‑resolution. Unlike methods that demand paired high‑resolution references—which are seldom accessible—our approach exclusively uses open‑source data. We start by applying KernelGAN to estimate the actual degradation kernel from each image. With this estimated kernel and a controlled noise injection, we generate realistic HR‑LR training pairs. These pairs are then used to train DeHAT‑SR, a specialized network built on a hierarchical attention Transformer. The hybrid attention and cross‑aggregation modules within the network allow it to exploit a much larger set of informative pixels during reconstruction. As a result, we avoid the training instability and spurious textures that often accompany GAN‑based methods. Our experiments on the SEN12MS dataset confirm that KDeHAT‑SR successfully enhances 10 m Sentinel‑2 imagery to 2.5 m resolution. When compared with BiCubic, EDSR, RCAN, ESRGAN, and Real‑ESRGAN, our model consistently achieves the best scores on the no‑reference metrics NIQE and BRISQUE, and the output images show noticeably sharper edges and more believable textures.

Tao Bian, Huibo Wang, Wenhui Xia, Wenjiao Deng, Yunhe Li. Super-Resolution Analysis of Sentinel-2 Remote Sensing Imagery Based on Degradation Kernel Estimation and Hierarchical Attention Transformer. International Journal of Multimedia Computing (2026), Vol. 7, Issue 2: 17-27. https://doi.org/10.38007/IJMC.2026.070203

[1] Bei Ye, Shufang Tian, Jia Ge, et al. 2017. Assessment of WorldView-3 data for lithological mapping. Remote Sensing 9, 11 (2017), 1132,19 pages. Darrel L. Williams, Samuel Goward, and Terry Arvidson. 2006. Landsat. Photogrammetric Engineering & Remote Sensing 72, 10 (2006), 1171–1178.

[2]  European Space Agency (ESA). 2015. Sentinel-2 User Handbook. Retrieved March 18, 2024 from https://earth.esa.int/documents/247904/685211/Sentinel-2_User_Handbook

[3] Massimiliano Gargiulo, Antonio Mazza, Raffaele Gaetano, et al. 2019. Fast Super-Resolution of 20 m Sentinel-2 Bands Using Convolutional Neural Networks. Remote Sensing 11, 22 (2019), 2635,18 pages.

[4] Massimiliano Gargiulo, Antonio Mazza, Raffaele Gaetano, et al. 2018. A CNN-based fusion method for super-resolution of Sentinel-2 data. In Proceedings of the IEEE International Geoscience and Remote Sensing Symposium (IGARSS '18). IEEE, 4713–4716.

[5] Luka Rumora, Mateo Gašparović, Mario Miler, et al. 2020. Quality assessment of fusing Sentinel-2 and WorldView-4 imagery on Sentinel-2 spectral band values. International Journal of Image and Data Fusion 11, 1 (2020), 77–96.

[6] Jianchao Yang, John Wright, Thomas S. Huang, et al. 2010. Image super-resolution via sparse representation. IEEE Transactions on Image Processing 19, 11 (2010), 2861–2873.

[7] Shuiping Gou, Shuzhen Liu, Shuyuan Yang, et al. 2014. Remote sensing image super-resolution reconstruction based on nonlocal pairwise dictionaries and double regularization. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing 7, 12 (2014), 4784–4792.

[8] Zongxu Pan, Jian Yu, Huijuan Huang, et al. 2013. Super-resolution based on compressive sensing and structural self-similarity for remote sensing images. IEEE Transactions on Geoscience and Remote Sensing 51, 9 (2013), 4864–4876.

[9] Chao Dong, Chen Change Loy, Kaiming He, et al. 2016. Image super-resolution using deep convolutional networks. IEEE Transactions on Pattern Analysis and Machine Intelligence 38, 2 (2016), 295–307.

[10] Yulun Zhang, Kunpeng Li, Kai Li, et al. 2018. Image super-resolution using very deep residual channel attention networks. In Proceedings of the European Conference on Computer Vision (ECCV '18). Springer, 294–310.

[11] Ian Goodfellow, Jean Pouget-Abadie, Mehdi Mirza, et al. 2020. Generative adversarial networks. Communications of the ACM 63, 11 (2020), 139–144.

[12] Christian Ledig, Lucas Theis, Ferenc Huszar, et al. 2017. Photo-realistic single image super-resolution using a generative adversarial network. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR '17). IEEE, 105–114.

[13] Xintao Wang, Ke Yu, Shixiang Wu, et al. 2018. ESRGAN: Enhanced super-resolution generative adversarial networks. In Proceedings of the European Conference on Computer Vision Workshops (ECCVW '18). Springer, 63–79.

[14] Jingyun Liang, Jiezhang Cao, Guolei Sun, et al. 2021. SwinIR: Image restoration using Swin Transformer. In Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV '21). IEEE, 1833–1844.

[15] Syed Waqas Zamir, Aditya Arora, Salman Khan, et al. 2022. Restormer: Efficient Transformer for high-resolution image restoration. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR '22). IEEE, 5718–5729.

[16] Marcos V. Conde, Kerem Turgutlu, Radu Timofte, et al. 2022. Swin2SR: SwinV2 Transformer for compressed image super-resolution and restoration. In Proceedings of the European Conference on Computer Vision Workshops (ECCVW '22). Springer, 669–687.

[17] Xiangyu Chen, Xintao Wang, Jiantao Zhou, et al. 2023. Activating more pixels in image super-resolution Transformer. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR '23). IEEE, 22367–22377.

[18] Zheng Chen, Yulun Zhang, Jinlin Gu, et al. 2023. Dual aggregation Transformer for image super-resolution. In Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV '23). IEEE, 12278–12287.

[19] Xintao Wang, Liangbin Xie, Chao Dong, et al. 2021. Real-ESRGAN: Training real-world blind super-resolution with pure synthetic data. In Proceedings of the IEEE/CVF International Conference on Computer Vision Workshops (ICCVW '21). IEEE, 1905–1914.

[20] Wenhui Ma, Zhaoqin Pan, Jian Guo, et al. 2018. Super-resolution of remote sensing images based on transferred generative adversarial network. In Proceedings of the IEEE International Geoscience and Remote Sensing Symposium (IGARSS '18). IEEE, 1148–1151.

[21] Yuan Gong, Pinghui Liao, Xiaoxin Zhang, et al. 2021. Enlighten-GAN for super resolution reconstruction in mid-resolution remote sensing images. Remote Sensing 13, 6 (2021), 1104,16 pages.

[22] Mikel Galar, Rubén Sesma, Christian Ayala, et al. 2020. Super-resolution of Sentinel-2 images using convolutional neural networks and real ground truth data. Remote Sensing 12, 18 (2020), 2941, 37 pages.

[23] Luis Salgueiro Romero, Javier Marcello, and Veronica Vilaplana. 2020. Super-resolution of Sentinel-2 imagery using generative adversarial networks. Remote Sensing 12, 15 (2020), 2424, 25 pages.

[24] Yunhe Li and Bo Li. 2021. Super-resolution of Sentinel-2 images at 10m resolution without reference images. Preprints (2021). https://doi.org/10.20944/preprints202104.0556.v1

[25] Chitwan Saharia, Jonathan Ho, William Chan, et al. 2022. Image super-resolution via iterative refinement. IEEE Transactions on Pattern Analysis and Machine Intelligence 45, 4 (2022), 4713–4726.

[26] Haoying Li, Yifan Yang, Meng Chang, et al. 2022. SRDiff: Single image super-resolution with diffusion probabilistic models. Neurocomputing 479 (2022), 47–59.

[27] Cesar Aybar, Julio Contreras, Simon Donike, Enrique Portalés-Julià, Gonzalo Mateo-García, and Luis Gómez-Chova. 2025. A Radiometrically and Spatially Consistent Super-Resolution Framework for Sentinel-2. SSRN.

[28] Sefi Bell-Kligler, Assaf Shocher, and Michal Irani. 2019. Blind super-resolution kernel estimation using an internal-GAN. arXiv preprint arXiv:1909.06581 (2019). Retrieved from

[29] Assaf Shocher, Shai Bagon, Phillip Isola, et al. 2019. InGAN: Capturing and remapping the "DNA" of a natural image. arXiv preprint arXiv:1812.00231 (2019). Retrieved from

[30] Jingwen Chen, Jiawei Chen, Hongyang Chao, et al. 2018. Image blind denoising with generative adversarial network based noise modeling. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR '18). IEEE, 3155–3164.

[31] Ruofan Zhou and Sabine Süsstrunk. 2019. Kernel modeling super-resolution on real low-resolution images. In Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV '19). IEEE, 2433–2443.

[32] Karen Simonyan and Andrew Zisserman. 2015. Very deep convolutional networks for large-scale image recognition. arXiv preprint arXiv:1409.1556 (2015). Retrieved from

[33] Michael Schmitt, Lloyd H. Hughes, Chunping Qiu, et al. 2019. SEN12MS - A curated dataset of georeferenced multi-spectral sentinel-1/2 imagery for deep learning and data fusion. arXiv preprint arXiv:1906.07789 (2019).

[34] Anish Mittal, Rajiv Soundararajan, and Alan C. Bovik. 2012. Making a "completely blind" image quality analyzer. IEEE Signal Processing Letters 20, 3 (2012), 209–212.

[35] Anish Mittal, Anush Krishna Moorthy, and Alan C. Bovik. 2012. No-reference image quality assessment in the spatial domain. IEEE Transactions on Image Processing 21, 12 (2012), 4695–4708.