MODE-GAN: Multi-order dual-edge generative adversarial network for remote sensing image super-resolution
Journal of King Saud University Computer and Information Sciences · 2026-07 · 38 (5)
DOI: 10.1007/s44443-026-00614-9 ↗Kurz gefasst
Fachartikel in der Zeitschrift Journal of King Saud University Computer and Information Sciences, online erschienen am 04.03.2026. Verfasst von Chunyang Wang und 5 weiteren Autor:innen. Die Arbeit nennt 54 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY-NC-ND 4.0). Gefördert von: Henan Provincial Science and Technology Research Project, Henan Polytechnic University for the Double First-Class Project of Surveying and Mapping Disciplines, Deep Earth Probe and Mineral Resources Exploration – National Science and Technology Major Project.
Abstract (Originalsprache)
The spatial resolution of remote sensing imagery is intrinsically limited by satellite sensor constraints and atmospheric degradation. To overcome these challenges, we propose the Multi-Order Dual-Edge Generative Adversarial Network (MODE-GAN). Our framework incorporates a novel Edge-Aware Feature Module (EAFM), which integrates first-order Scharr gradients with second-order multi-scale Laplacian of Gaussian (mLoG) operators as physics-inspired guidance. To ensure physical consistency for quantitative analysis, we employ the Spectral Angle Mapper (SAM) as a critical validation metric to assess radiometric fidelity. We validated the model using 12,996 real-world, cross-sensor pairs from Landsat-8 and Sentinel-2. Experimental results demonstrate that MODE-GAN significantly outperforms state-of-the-art architectures. Quantitatively, the model achieves a PSNR of 24.30 dB and an SSIM of 0.77, outperforming the second-best model by 0.76 dB and 0.03, respectively. Notably, MODE-GAN preserves the intrinsic spectral shape with a SAM value of 7.48, superior to HAT (8.53) and Real-ESRGAN (8.11). Stress tests confirm high resilience to sub-pixel registration noise, ensuring structural integrity in non-aligned scenarios. By integrating multi-order differential physical operators as internal feature guidance and structural loss constraints, MODE-GAN achieves enhanced geometric structure and spectral fidelity, providing a reliable pathway for downstream quantitative applications such as NDVI estimation and land-cover mapping.
Förderung
- Henan Provincial Science and Technology Research Project
- Henan Polytechnic University for the Double First-Class Project of Surveying and Mapping Disciplines
- Deep Earth Probe and Mineral Resources Exploration – National Science and Technology Major Project
Zitieren
APA
Wang, C., Zheng, L., Yang, W., Gu, Y., Wang, G., & Chen, Z. (2026). MODE-GAN: Multi-order dual-edge generative adversarial network for remote sensing image super-resolution. Journal of King Saud University Computer and Information Sciences, 38(5). https://doi.org/10.1007/s44443-026-00614-9
Vancouver
Wang C, Zheng L, Yang W, Gu Y, Wang G, Chen Z. MODE-GAN: Multi-order dual-edge generative adversarial network for remote sensing image super-resolution. Journal of King Saud University Computer and Information Sciences. 2026;38(5). doi:10.1007/s44443-026-00614-9
BibTeX
@article{Wang2026,
author = {Wang, Chunyang and Zheng, Licheng and Yang, Wei and Gu, Yanan and Wang, Gaige and Chen, Zhichao},
title = {MODE-GAN: Multi-order dual-edge generative adversarial network for remote sensing image super-resolution},
journal = {Journal of King Saud University Computer and Information Sciences},
year = {2026},
volume = {38},
number = {5},
doi = {10.1007/s44443-026-00614-9}
}
EUVIMED