Frequency-Enhanced Feature Pyramid Network With Global Saliency Kernel Module for Infrared Small Target Detection
IEEE Signal Processing Letters · 2026 · 33 · 2650-2654
DOI: 10.1109/lsp.2026.3705382 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Signal Processing Letters, 2026. Verfasst von Simiao Wang und 2 weiteren Autor:innen. Die Arbeit nennt 36 Literaturangaben. Gefördert von: Taishan Scholar, National Natural Science Foundation of China.
Abstract (Originalsprache)
Infrared Small Target Detection (ISTD) plays a critical role in many civil and military applications due to its reliable all-day imaging capability. However, detecting extremely small targets under low signal-to-clutter ratio conditions remains highly challenging. Existing deep models rely on relatively heavy network architectures, yet fail to capture feature representations that are sensitive to small targets. To address these issues, this work proposes a Frequency-Enhanced Feature Pyramid Network (FE-FPN), which simultaneously strengthens spatial saliency representation and frequency-domain representation to facilitate ISTD. Specifically, a Global Saliency Kernel Module (GSKM) introduces a pixel-level adaptive enhancement mechanism inspired by Human Visual System priors to improve local saliency modeling. In addition, a Frequency Mutual Learning Module (FMLM) fully exploits the complementary properties of amplitude and phase to enrich high-frequency details while preserving structural information. Extensive experiments on benchmark datasets demonstrate that the proposed method outperforms state-of-the-art approaches, and ablation studies further verify the effectiveness of GSKM and FMLM in improving robustness and detection accuracy.
Förderung
- Taishan Scholar
- National Natural Science Foundation of China
Zitieren
APA
Wang, S., Liu, Y., & Lu, M. (2026). Frequency-Enhanced Feature Pyramid Network With Global Saliency Kernel Module for Infrared Small Target Detection. IEEE Signal Processing Letters, 33, 2650-2654. https://doi.org/10.1109/lsp.2026.3705382
Vancouver
Wang S, Liu Y, Lu M. Frequency-Enhanced Feature Pyramid Network With Global Saliency Kernel Module for Infrared Small Target Detection. IEEE Signal Processing Letters. 2026;33:2650-2654. doi:10.1109/lsp.2026.3705382
BibTeX
@article{Wang2026,
author = {Wang, Simiao and Liu, Yunan and Lu, Mingyu},
title = {Frequency-Enhanced Feature Pyramid Network With Global Saliency Kernel Module for Infrared Small Target Detection},
journal = {IEEE Signal Processing Letters},
year = {2026},
volume = {33},
pages = {2650--2654},
doi = {10.1109/lsp.2026.3705382}
}
EUVIMED