Effective Linear-Array Computational Hyperspectral Imaging With Multilayer Thin Films

Spectral imaging technology has extensive applications in remote sensing, medical diagnosis, biomedical engineering, archaeology and art conservation, and food testing. Conventional cameras use 4 channels (RGGB) to reconstruct color images. Our system employs 4 broadband-encoding random filters as encoders and utilizes a multi-level spectral transformer (MST) as the software decoder. This enables the reconstruction of hyperspectral images across 61 channels ranging from 410 nm to 890 nm, at a resolution of $1280\times 1080$ . Experimental results demonstrate that our framework can accurately reconstruct hyperspectral images in both indoor and outdoor settings. This technology has the potential to replace existing color-imaging sensors and shows promising prospects in the fields of imaging and spectral analysis.