Spectral selectivity in compressive spectral imaging based on grayscale coded apertures

Abstract

Compressive Spectral Imaging (CSI) is a signal acquisition technique that captures a spatial map of the spectral variation of a scene. Recently, a new optical imaging architecture called Coded Aperture Snapshot Spectral Imaging (CASSI) has emerged. The CASSI emulates the role of a spectrometer insomuch that it captures spectral information but uses coded apertures to take 2D compressed measurements from a 3D scene. Subsequently, an optimization algorithm is used to recover the full 3D spectral image from the measurements. However, in some applications is required to recover just a few selected set of spectral information. Then, compressive spectral selectivity aims to recover a specific set of spectral bands of interest. This work extends the capabilities of CASSI by replacing the traditional block-unblock coded apertures for a grayscale valued coded aperture. Further, the structures of the gray scale-valued coded apertures are designed such that a specific set of selected bands is reconstructed with high quality. A forward model and its corresponding reconstruction method are presented allowing to recover the desired bands, exclusively. Simulations are performed obtaining reconstructions exhibiting PSNRs of up to 30 dB.