Hyperspectral compressive microscopy based on structured light sheet and deep convolutional neural network
Résumé
We describe a compressive hyperspectral microscope based on a structured light sheet and a deep convolutional neural network. The setup extends the concept of the computational hyperspectral microscope introduced in Ref. 1. First, we significantly improve the quality of the structured light sheet using a digital micro-mirror device to generate the light patterns directly. Second, to reduce the acquisition time, only fewer light patterns are acquired. The resulting inverse problem is solved using a deep neural network that includes traditional Tikhonov regularization.