EUVIMEDEUVIMED entdecken
Crossref

AIMS-Net: An Adaptive Intermediate Feature Sharing Framework for Multispectral Image Compression

Qizhi Fang, Yuxuan Duan, Jingang Wang, Lili Zhang

IEEE Transactions on Aerospace and Electronic Systems · 2026 · 62 · 19099-19114

DOI: 10.1109/taes.2026.3740375 ↗

Kurz gefasst

Fachartikel in der Zeitschrift IEEE Transactions on Aerospace and Electronic Systems, 2026. Verfasst von Qizhi Fang und 3 weiteren Autor:innen. Die Arbeit nennt 49 Literaturangaben. Gefördert von: Fundamental Research Funds for Universities of Liaoning Province.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Existing deep learning-based compression models primarily focus on learning latent representations, often treating the encoding–decoding process as an opaque process and neglecting the potential of intermediate features to provide semantic priors for reconstruction. To address these limitations, this article proposes the adaptive intermediate multispectral sharing network (AIMS-Net) that employs an intermediate-layer semantic prototype sharing mechanism. In this mechanism, a latent feature bridge transmits low-entropy semantic priors between the encoder and decoder. This process adaptively modulates the probability distribution of latent representations. Furthermore, an Adaptive Spectral–Context Modulation module is designed to suppress interspectral redundancy. It enhances spectral discriminability through a dual-stream interaction mechanism. In parallel, a Progressive Sparse Context Aggregation module is proposed. This module efficiently aggregates key structural information. It reduces computational overhead through multiscale modeling and sparse attention mechanisms. Experimental results demonstrate that AIMS-Net outperforms state-of-the-art compression methods across all evaluated metrics. Specifically, at a bit rate of approximately 0.2 bits per pixel per band, AIMS-Net achieves a reconstruction peak signal-to-noise ratio (PSNR) of 60.79 dB on the Sentinel-2 test set. These results demonstrate the strong potential of the proposed method for efficient compression of multispectral remote sensing images.

Förderung

  • Fundamental Research Funds for Universities of Liaoning Province

Zitieren

APA

Fang, Q., Duan, Y., Wang, J., & Zhang, L. (2026). AIMS-Net: An Adaptive Intermediate Feature Sharing Framework for Multispectral Image Compression. IEEE Transactions on Aerospace and Electronic Systems, 62, 19099-19114. https://doi.org/10.1109/taes.2026.3740375

Vancouver

Fang Q, Duan Y, Wang J, Zhang L. AIMS-Net: An Adaptive Intermediate Feature Sharing Framework for Multispectral Image Compression. IEEE Transactions on Aerospace and Electronic Systems. 2026;62:19099-19114. doi:10.1109/taes.2026.3740375

BibTeX
@article{Fang2026,
  author = {Fang, Qizhi and Duan, Yuxuan and Wang, Jingang and Zhang, Lili},
  title = {AIMS-Net: An Adaptive Intermediate Feature Sharing Framework for Multispectral Image Compression},
  journal = {IEEE Transactions on Aerospace and Electronic Systems},
  year = {2026},
  volume = {62},
  pages = {19099--19114},
  doi = {10.1109/taes.2026.3740375}
}
Kostenlos starten →