EUVIMEDEUVIMED entdecken
Crossref

DULRNet: A Deep Unfolding Low-Rank Network for Infrared and Visible Image Fusion

Dongming Wang, Xingang Mou, Ze Huang, Xiao Zhou

IEEE Sensors Journal · 15.07.2026 · 26 (14) · 21508-21526

DOI: 10.1109/jsen.2026.3700714 ↗

Kurz gefasst

Fachartikel in der Zeitschrift IEEE Sensors Journal, erschienen am 15.07.2026. Verfasst von Dongming Wang und 3 weiteren Autor:innen. Die Arbeit nennt 58 Literaturangaben. Gefördert von: National Natural Science Foundation of China.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Infrared and visible image fusion (IVIF) aims to integrate prominent thermal radiation targets with rich visible texture details. However, existing deep learning methods, which typically rely on end-to-end networks that directly mix features, lack an interpretable mathematical decomposition process to specifically guide feature fusion, and are ineffective in preserving the distinct advantages of both infrared and visible modalities. To address the issue, we propose a deep unfolding low-rank network (DULRNet), which explicitly unfolds the iterative optimization process of low-rank sparse decomposition (LRSD) into a multiscale dual-branch architecture. To alleviate the cross-scale noise propagation inherent in the unfolding process, a context-aware gated fusion module (CGFM) is designed to dynamically filter redundant features via spatially adaptive masks. Considering the physical asymmetry between thermal radiation and visible textures within the unfolding framework, an asymmetric bidirectional cross-modal attention (ABCA) mechanism is introduced to achieve differentiated feature fusion. Meanwhile, a dynamic coarse-to-fine training strategy is employed to balance structural and textural optimization. Extensive experiments on the MSRS, TNO, and RoadScene datasets demonstrate that DULRNet achieves state-of-the-art performance in both visual fidelity and objective metrics. Furthermore, it significantly enhances robustness in downstream vision tasks, including object detection and semantic segmentation. The code is available at https://github.com/wang2025-source/DULRNet

Förderung

  • National Natural Science Foundation of China

Zitieren

APA

Wang, D., Mou, X., Huang, Z., & Zhou, X. (2026). DULRNet: A Deep Unfolding Low-Rank Network for Infrared and Visible Image Fusion. IEEE Sensors Journal, 26(14), 21508-21526. https://doi.org/10.1109/jsen.2026.3700714

Vancouver

Wang D, Mou X, Huang Z, Zhou X. DULRNet: A Deep Unfolding Low-Rank Network for Infrared and Visible Image Fusion. IEEE Sensors Journal. 2026;26(14):21508-21526. doi:10.1109/jsen.2026.3700714

BibTeX
@article{Wang2026,
  author = {Wang, Dongming and Mou, Xingang and Huang, Ze and Zhou, Xiao},
  title = {DULRNet: A Deep Unfolding Low-Rank Network for Infrared and Visible Image Fusion},
  journal = {IEEE Sensors Journal},
  year = {2026},
  volume = {26},
  number = {14},
  pages = {21508--21526},
  doi = {10.1109/jsen.2026.3700714}
}
Kostenlos starten →