P.O. Arkhipov1, S.L. Philippskih2, M.V. Tsukanov3

1–3 Orel Branch of the Federal Research Center "Computer Science and Control" of the Russian Academy of Sciences (Orel, Russia)
1 arpaul@mail.ru, 2 philippsl@mail.ru, 3 tsukanov.m.v@yandex.ru

This paper examines the limitations of existing methods for aerial and satellite image super-resolution. Such methods are primarily designed to improve the visual quality of images but often introduce inaccuracies and hallucinations, limiting their applicability to computer vision and Earth observation tasks. To overcome these limitations, a neural network training method capable of controlling hallucinations is required to ensure that the reconstructed images remain as close as possible to the corresponding ground-truth high-resolution images. To address this challenge, a study was conducted to identify baseline models for evaluating the effectiveness of the proposed loss function and the training protocol for convolutional neural networks designed for aerial and satellite image super-resolution. The study established a baseline reconstruction performance using bicubic interpolation, identified SRResNet as the most suitable neural network architecture, and developed a training protocol incorporating both synthetic and cross-sensor satellite imagery. The obtained results provide the foundation for developing a neural network training method with hallucination control and a dedicated loss function aimed at preserving the natural details of aerial and satellite images.

Arkhipov P.O., Philippskih S.L., Tsukanov M.V. Investigation of baseline neural network architectures and a training protocol for convolutional neural network models to improve the quality of aerial images // Highly Available Systems. 2026. V. 22. № 3. P. 45−54. DOI: https://doi.org/10.18127/j20729472-202603-05

1. Wang Z., Chen J., Hoi S.C.H. Deep Learning for Image Super-Resolution: A Survey. IEEE Transactions on Pattern Analysis and Machine Intelligence. 2021. V. 43. № 10. P. 3365–3387. doi: 10.1109/TPAMI.2020.2982166
2. Arkhipov P.O., Philippskikh S.L., Tsukanov M.V. Osobennosti prakticheskoj realizacii metoda augmentacii malyh ob’ektov na osnove setej sverhvysokogo razresheniya. Sistemy i Sredstva Informatiki. 2025. V. 35. № 3. P. 17–32. doi: 10.14357/08696527250302 (in Russian).
3. Arkhipov P.O., Philippskikh S.L., Tsukanov M.V. Development of a Specialized Method for Contextual Augmentation of Small Objects. Pattern Recognition and Image Analysis. 2026. V. 36. № 2. P. 335–342. doi: 10.1134/S1054661826700148
4. Dong C., Loy C.C., Tang X. Accelerating the Super-Resolution Convolutional Neural Network. Computer Vision – ECCV 2016. Lecture Notes in Computer Science. Cham: Springer, 2016. V. 9906. P. 391–407. doi: 10.1007/978-3-319-46475-6_25
5. Zhang Y., Li K., Li K., Wang L., Zhong B., Fu Y. Image Super-Resolution Using Very Deep Residual Channel Attention Networks. Computer Vision – ECCV 2018. Lecture Notes in Computer Science. Cham: Springer, 2018. V. 11211. P. 294–310. doi: 10.1007/978-3-030-01234-2_18
6. Ledig C., Theis L., Huszár F., Caballero J., Cunningham A., Acosta A., Aitken A., Tejani A., Totz J., Wang Z., Sh W. Photo-Realistic Single Image Super-Resolution Using a Generative Adversarial Network. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR). 2017. P. 4681–4690. doi: 10.1109/CVPR.2017.19
7. Wang X., Yu K., Wu S., Gu J., Liu Y., Dong C., Qiao Y., Loy C.C. ESRGAN: Enhanced Super-Resolution Generative Adversarial Networks. Computer Vision – ECCV 2018 Workshops. Lecture Notes in Computer Science. Cham: Springer, 2019. V. 11133. P. 63–79. doi: 10.1007/978-3-030-11021-5_5
8. Wang X., Xie L., Dong C., Shan Y. Real-ESRGAN: Training Real-World Blind Super-Resolution with Pure Synthetic Data. Proceedings
   of the IEEE/CVF International Conference on Computer Vision Workshops (ICCVW). 2021. P. 1905–1914. doi: 10.1109/ICCVW54120.2021.00217
9. Liang J., Cao J., Sun G., Zhang K., Van Gool L., Timofte R. SwinIR: Image Restoration Using Swin Transformer. Proceedings of the IEEE/CVF International Conference on Computer Vision Workshops (ICCVW). 2021. P. 1833–1844. doi: 10.1109/ICCVW54120.2021.00210
10. Zhang A., Lipton Z.C., Li M., Smola A.J. Dive into Deep Learning. Cambridge: Cambridge University Press, 2023. 842 p. ISBN: 978-1-009-38294-5
11. Chen H., He X., Qing L., Wu Y., Ren C., Sheriff R.E., Zhu C. Real-world Single Image Super-Resolution: A Brief Review. Information Fusion. 2022. V. 79. P. 124–145. doi: 10.1016/j.inffus.2021.09.005
12. Donike S., Aybar C., Gómez-Chova L., Kalaitzis F. Trustworthy Super-Resolution of Multispectral Sentinel-2 Imagery With Latent Diffusion. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing. 2025. doi: 10.1109/JSTARS.2025.3542220
13. Ren W., Goyal R., Hu Z. Aumentado-Armstrong T.T., Mohomed I., Levinshtein A. Hallucination Score: Towards Mitigating Hallucinations in Generative Image Super-Resolution. arXiv:2507.14367 [cs.CV]. 2025. doi: 10.48550/arXiv.2507.14367
14. ESA OpenSR Homepage. URL: https://esaopensr.github.io/ (data obrashcheniya: 21.07.2026).
15. Drusch M., Del Bello U., Carlier S. et al. Sentinel-2: ESA's Optical High-Resolution Mission for GMES. Remote Sensing of Environment. 2012. V. 120. P. 25–36. doi: 10.1016/j.rse.2011.11.026
16. Aybar C., Montero D., Contreras J., Donike S., Kalaitzis F., Gómez-Chova L. SEN2NAIP: A Large-Scale Dataset for Sentinel-2 Image Super-Resolution. Scientific Data. 2024. V. 11. Article 1389. doi: 10.1038/s41597-024-04214-y
17. Aybar C., Montero D., Donike S., Kalaitzis F., Gómez-Chova L. A Comprehensive Benchmark for Optical Remote Sensing Image Super-Resolution. IEEE Geoscience and Remote Sensing Letters. 2024. V. 21. Art. 5003105. P. 1–5. doi: 10.1109/LGRS.2024.3401394
18. Donike S., Aybar C., Contreras J., Gómez-Chova L. OpenSR-SRGAN: A Flexible Super-Resolution Framework for Multispectral Earth Observation Data. arXiv:2511.10461 [cs.CV]. 2025. doi: 10.48550/arXiv.2511.10461
19. Willmott C.J., Matsuura K. Advantages of the Mean Absolute Error (MAE) over the Root Mean Square Error (RMSE) in Assessing Average Model Performance. Climate Research. 2005. V. 30. № 1. P. 79–82. doi: 10.3354/cr030079
20. Kruse F.A., Lefkoff A.B., Boardman J.W., Heidebrecht K.B., Shapiro A.T., Barloon P.J., Goetz A.F.H. The Spectral Image Processing System (SIPS) – Interactive Visualization and Analysis of Imaging Spectrometer Data. Remote Sensing of Environment. 1993. V. 44.
    № 2–3. P. 145–163. doi: 10.1016/0034-4257(93)90013-N
21. Chang C.-I. Hyperspectral Imaging: Techniques for Spectral Detection and Classification. New York: Kluwer Academic. Plenum Publi­shers. 2003. 370 p. doi: 10.1007/978-1-4615-0285-9
22. Keys R.G. Cubic Convolution Interpolation for Digital Image Processing. IEEE Transactions on Acoustics, Speech, and Signal Processing. 1981. V. 29. № 6. P. 1153–1160. doi: 10.1109/TASSP.1981.1163711
23. Howse J., Minichino J. Learning OpenCV 5 Computer Vision with Python. 4th ed. Birmingham: Packt Publishing. 2023.
24. CKP "INFORMATIKA". URL: https://www.frccsc.ru/ckp (data obrashcheniya: 21.07.2026). (in Russian).