EUVIMEDEUVIMED entdecken
Crossref

DDPM-Driven Background Suppression for Efficient and Accurate Hyperspectral Anomaly Change Detection

Xing Hu, Houwei Liu, Danfeng Hong, Qianqian Duan, Dawei Zhang

IEEE Transactions on Circuits and Systems for Video Technology · 2026-11 · 36 (11) · 16438-16453

DOI: 10.1109/tcsvt.2026.3718648 ↗

Kurz gefasst

Fachartikel in der Zeitschrift IEEE Transactions on Circuits and Systems for Video Technology, 2026-11. Verfasst von Xing Hu und 4 weiteren Autor:innen. Die Arbeit wurde bisher 1-mal zitiert (Crossref) und nennt 47 Literaturangaben. Gefördert von: High-Quality Development Special Project of the Ministry of Industry and Information Technology, Academician Workstation Program of Yunnan Province, Shanghai Agricultural Technology Innovation Project.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Hyperspectral anomaly change detection aims to precisely capture the subtle spectral differences between multitemporal hyperspectral images (HSIs) in order to accurately identify changes on the Earth’s surface. However, this field faces numerous challenges. On one hand, differences in imaging conditions of multitemporal HSIs lead to the confusion between real surface changes and pseudo changes caused by variations in imaging conditions. On the other hand, significant background differences in multitemporal HSIs result in poor performance of traditional background suppression methods. In light of these challenges, this paper applies Denoising Diffusion Probabilistic Models (DDPMs) to the field of background suppression in hyperspectral anomaly change detection and proposes a novel unsupervised hyperspectral change detection algorithm—DDPM-Driven Background Suppression Method (DDBS). DDBS employs a background reconstruction-suppression hybrid processing structure. It reduces the background differences between multi-temporal HSIs through the background reconstruction process and simulates the common background of bi-temporal hyperspectral images through a novel DDPMs noise generation method. During the inference process, the common background is removed as noise, thereby suppressing the common background in multi-temporal HSIs and highlighting anomalous changes. Furthermore, the proposed hybrid background reconstruction-suppression architecture only requires a single inference step to effectively suppress common background components in multitemporal HSIs. Experimental results on multiple public datasets demonstrate that the proposed DDBS outperforms state-of-the-art background reconstruction and suppression methods in hyperspectral anomaly change detection tasks.

Förderung

  • High-Quality Development Special Project of the Ministry of Industry and Information Technology
  • Academician Workstation Program of Yunnan Province
  • Shanghai Agricultural Technology Innovation Project

Zitieren

APA

Hu, X., Liu, H., Hong, D., Duan, Q., & Zhang, D. (2026). DDPM-Driven Background Suppression for Efficient and Accurate Hyperspectral Anomaly Change Detection. IEEE Transactions on Circuits and Systems for Video Technology, 36(11), 16438-16453. https://doi.org/10.1109/tcsvt.2026.3718648

Vancouver

Hu X, Liu H, Hong D, Duan Q, Zhang D. DDPM-Driven Background Suppression for Efficient and Accurate Hyperspectral Anomaly Change Detection. IEEE Transactions on Circuits and Systems for Video Technology. 2026;36(11):16438-16453. doi:10.1109/tcsvt.2026.3718648

BibTeX
@article{Hu2026,
  author = {Hu, Xing and Liu, Houwei and Hong, Danfeng and Duan, Qianqian and Zhang, Dawei},
  title = {DDPM-Driven Background Suppression for Efficient and Accurate Hyperspectral Anomaly Change Detection},
  journal = {IEEE Transactions on Circuits and Systems for Video Technology},
  year = {2026},
  volume = {36},
  number = {11},
  pages = {16438--16453},
  doi = {10.1109/tcsvt.2026.3718648}
}
Kostenlos starten →