Establish broader design principles for metasurface-based snapshot hyperspectral imaging

Establish broader design principles for metasurface-based snapshot hyperspectral imaging, including how datacube sampling, reconstruction priors, spatial–spectral mixing strategies, metasurface parametrizations, and reconstruction algorithms jointly determine imaging performance.

Background

The paper studies snapshot hyperspectral imaging systems that encode a three-dimensional spatial–spectral datacube onto a two-dimensional detector using a metasurface and computational reconstruction. Performance depends jointly on the datacube’s spatial–spectral sampling regime, the strength and form of the reconstruction prior, the optical encoding strategy, the physical parametrization of the metasurface, and the reconstruction algorithm.

The authors identify a gap in general design principles because strong reconstruction results may arise either from informative optical measurements or from a restrictive learned prior. Their experiments show that simple single-parameter dielectric-pillar metasurfaces often favor lens-like spatial encoding with weak spectral discrimination, while end-to-end reconstruction can conceal this optical limitation on restricted datasets. The unresolved problem is therefore to determine general principles governing the co-design and allocation of spatial and spectral information in physically realizable metasurface-based snapshot hyperspectral imagers.