Modality Complementarity Meets Image Enhancement: A Visible--Infrared Collaborative Enhancement Framework for Full-Time AAVs Perception
IEEE Transactions on Circuits and Systems for Video Technology · 2026-11 · 36 (11) · 17228-17245
DOI: 10.1109/tcsvt.2026.3716488 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Transactions on Circuits and Systems for Video Technology, 2026-11. Verfasst von Huan Wang und 4 weiteren Autor:innen. Die Arbeit nennt 72 Literaturangaben. Gefördert von: National Natural Science Foundation of China, China Postdoctoral Science Foundation, Program for Science and Technology Innovation Talents in Universities of Henan Province.
Abstract (Originalsprache)
Multi-modal image collaborative enhancement in autonomous aerial vehicles (AAVs) enables the complementary integration of fine-grained visible details and thermal cues, thereby enhancing robust scene perception under adverse conditions. However, when modal information becomes imbalanced or cross-modal interference occurs, existing single-modal enhancement and guided super-resolution methods often introduce artifacts and undermine structural fidelity. To address these issues, we propose a Visible–Infrared Dual-Branch Collaborative Enhancement Network (VICENet). Specifically, the visible light (VIS) branch integrates a Frequency Adaptive Fusion Block (FAFB), which employs a learnable frequency mask to perform degradation-aware spectral decomposition. This design adaptively identifies degraded frequency bands and jointly restores textures and structures via a frequency-guided refinement mechanism. The infrared (IR) branch enhances thermal representations via long-range relation modeling and global context reasoning, achieving stable recovery under complex conditions. Moreover, we introduce a Multi-modal Feature Interaction Module (MFIM), which performs joint spatial-frequency analysis through wavelet transforms and adopts a frequency-aware gating strategy for selective cross-modal feature fusion. To further enhance this interaction, the module employs a bidirectional adaptive attention mechanism that captures long-range dependencies, thereby achieving semantic alignment and robust cross-modal collaboration. Extensive experiments on public benchmarks demonstrate that VICENet consistently outperforms recent state-of-the-art baselines across both VIS enhancement and IR super-resolution tasks, highlighting its superior restoration capability and strong generalization for full-time AAVs perception.
Förderung
- National Natural Science Foundation of China
- China Postdoctoral Science Foundation
- Program for Science and Technology Innovation Talents in Universities of Henan Province
- Natural Science Foundation of Henan
- Training Program for Young Backbone Teachers in Higher Education Institutions of Henan Province
- Key Technological Breakthrough Projects for Major Industries in Henan Province
- National Key Research and Development Program of China
- Scientific and Technological Project of Henan Province
Zitieren
APA
Wang, H., Ma, J., Cui, W., Ji, B., & Si, P. (2026). Modality Complementarity Meets Image Enhancement: A Visible--Infrared Collaborative Enhancement Framework for Full-Time AAVs Perception. IEEE Transactions on Circuits and Systems for Video Technology, 36(11), 17228-17245. https://doi.org/10.1109/tcsvt.2026.3716488
Vancouver
Wang H, Ma J, Cui W, Ji B, Si P. Modality Complementarity Meets Image Enhancement: A Visible--Infrared Collaborative Enhancement Framework for Full-Time AAVs Perception. IEEE Transactions on Circuits and Systems for Video Technology. 2026;36(11):17228-17245. doi:10.1109/tcsvt.2026.3716488
BibTeX
@article{Wang2026,
author = {Wang, Huan and Ma, Jingbo and Cui, Wei and Ji, Baofeng and Si, Pengju},
title = {Modality Complementarity Meets Image Enhancement: A Visible--Infrared Collaborative Enhancement Framework for Full-Time AAVs Perception},
journal = {IEEE Transactions on Circuits and Systems for Video Technology},
year = {2026},
volume = {36},
number = {11},
pages = {17228--17245},
doi = {10.1109/tcsvt.2026.3716488}
}
EUVIMED