Hyperspectral Image Denoising via Deep Block Term Decomposition
IEEE Geoscience and Remote Sensing Letters · 2026 · 23 · 1-5
DOI: 10.1109/lgrs.2026.3678023 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Geoscience and Remote Sensing Letters, 2026. Verfasst von Zhirui Min und 2 weiteren Autor:innen. Die Arbeit nennt 20 Literaturangaben. Gefördert von: Sichuan Science and Technology Program, NSFC.
Abstract (Originalsprache)
Recently, block-term decomposition (BTD) has been successfully applied to hyperspectral image (HSI) processing. In particular, the $rank\text {-}(L,L,1)$ BTD, also known as the LL1 decomposition, decomposes the HSI tensor as a sum of component tensors formed by matrix–vector outer products of $rank\text {-}L$ core component matrices and vectors. However, these fundamental building blocks, i.e., core component matrices and matrix–vector outer products, exhibit limited flexibility, struggling to effectively capture the complex information in HSIs. To tackle this limitation, we propose a novel deep BTD (DBTD), which leverages attention-enhanced generators to generate core component tensors instead of core component matrices, and substitutes the fixed matrix–vector outer product with a learnable convolutional mechanism. This approach enables the proposed DBTD to achieve greater flexibility and expressive power for complex HSI structures. Notably, our proposed DBTD serves as a generalization of the classical LL1 decomposition, which degenerates to the classical LL1 decomposition when the core component tensor is simplified back to a low-rank matrix. Building upon the proposed DBTD, we propose a DBTD-based model for HSI denoising. To solve the resulting model, we develop an efficient alternating minimization-based algorithm. Extensive experimental results collectively demonstrate that the proposed method outperforms the competing methods.
Förderung
- Sichuan Science and Technology Program
- NSFC
Zitieren
APA
Min, Z., Huang, T., & Wu, W. (2026). Hyperspectral Image Denoising via Deep Block Term Decomposition. IEEE Geoscience and Remote Sensing Letters, 23, 1-5. https://doi.org/10.1109/lgrs.2026.3678023
Vancouver
Min Z, Huang T, Wu W. Hyperspectral Image Denoising via Deep Block Term Decomposition. IEEE Geoscience and Remote Sensing Letters. 2026;23:1-5. doi:10.1109/lgrs.2026.3678023
BibTeX
@article{Min2026,
author = {Min, Zhirui and Huang, Ting-Zhu and Wu, Wei-Hao},
title = {Hyperspectral Image Denoising via Deep Block Term Decomposition},
journal = {IEEE Geoscience and Remote Sensing Letters},
year = {2026},
volume = {23},
pages = {1--5},
doi = {10.1109/lgrs.2026.3678023}
}
EUVIMED