EUVIMEDEUVIMED entdecken
Crossref

Compact metasurface-based spatially multiplexed snapshot spectral imaging via joint learning of optical encoding and spectral reconstruction

Peng Lin, He Ren, Zihao Wan, Shouqian Chen

Optics Express · 16.11.2026 · 34 (23) · 44693

DOI: 10.1364/oe.616703 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Optics Express, online erschienen am 10.11.2026. Verfasst von Peng Lin und 3 weiteren Autor:innen. Die Arbeit nennt 39 Literaturangaben. Gefördert von: National Natural Science Foundation of China.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Compact snapshot spectral imaging is attractive for applications requiring high frame rates and compact instrumentation, yet existing systems often face limited spectral encoding diversity or a tradeoff between spectral sampling and spatial resolution. We introduce metasurface spectral encoding into a spatially multiplexed multi-aperture snapshot architecture and further develop a differentiable metasurface co-design (DMC) framework to jointly optimize the spectral encoding responses and spectral reconstruction. The optical architecture integrates a metasurface filter array, a lens array, and a monochrome sensor to simultaneously acquire multiple spectrally encoded subimages of the same scene in a single exposure. Distinct metasurface responses are assigned to separate sub-apertures, enabling spatially continuous acquisition with diverse broadband spectral encoding. A differentiable metasurface co-design (DMC) framework is developed to jointly determine the spectral encoding responses and learn the mapping from compressed measurements to the target spectrum. The candidate metasurface response library is incorporated as a differentiable optical encoding layer, together with a learnable multichannel selection module that automatically identifies a set of complementary responses according to the reconstruction objective. A temperature-controlled softmax relaxation enables the discrete response selection to be optimized together with the spectral reconstruction network, allowing the encoding configuration to adapt directly to the reconstruction. Numerical simulations on the CAVE and ICVL datasets demonstrate effective spectral reconstruction with the proposed DMC approach, achieving average PSNR values of 40.283 dB and 46.298 dB, respectively. These results indicate the effectiveness of the proposed framework for compact snapshot spectral imaging.

Förderung

  • National Natural Science Foundation of China

Zitieren

APA

Lin, P., Ren, H., Wan, Z., & Chen, S. (2026). Compact metasurface-based spatially multiplexed snapshot spectral imaging via joint learning of optical encoding and spectral reconstruction. Optics Express, 34(23), 44693. https://doi.org/10.1364/oe.616703

Vancouver

Lin P, Ren H, Wan Z, Chen S. Compact metasurface-based spatially multiplexed snapshot spectral imaging via joint learning of optical encoding and spectral reconstruction. Optics Express. 2026;34(23):44693. doi:10.1364/oe.616703

BibTeX
@article{Lin2026,
  author = {Lin, Peng and Ren, He and Wan, Zihao and Chen, Shouqian},
  title = {Compact metasurface-based spatially multiplexed snapshot spectral imaging via joint learning of optical encoding and spectral reconstruction},
  journal = {Optics Express},
  year = {2026},
  volume = {34},
  number = {23},
  pages = {44693},
  doi = {10.1364/oe.616703}
}
Kostenlos starten →