Grating-based coded aperture compressive spectral imaging to reconstruct over 190 spectral bands from a snapshot measurement

Abstract We report snapshot compressive spectral imaging using a grating and coded aperture. A vision-transformer-based deep learning algorithm is developed for high-quality reconstruction. Experimental results demonstrate that over 190 spectral bands can be reconstructed from a single measurement from 485–657 nm with spectral accuracy of about 2 nm and spectral resolution up to 1.5 nm.