Fast hyperspectral imaging via spectral undersampling using a digital micromirror device architecture

Hyperspectral imaging (HSI) captures spatial and spectral information from a scene. Conventional scanning HSI systems often yield low optical throughput, slow frame rates, large datasets and/or resolution trade-offs. One can exploit spectral compression in applications where spectral information is sparse in some representation and illumination is controllable, like optical endoscopy. This work experimentally demonstrates fast HSI using a compressed sensing architecture involving a programmable illuminator and a fast monochrome camera. We reconstruct the scene's hypercube, showing high PSNR (up to 39 for 10 measurements). Our versatile implementation offers higher optical throughput, smaller acquisition times and data volumes compared to conventional HSI approaches.