Sensing matrix design for spatially-selective compressive spectral imaging

Abstract

Compressive Spectral Imaging (CSI) is a feasible technique for capturing three-dimensional spatio-spectral information of a scene using fewer measurements than those required by traditional methods. The Coded Aperture Snapshot Spectral Imaging (CASSI) system is an optical architecture that exploits the CSI theory capturing the spectral information by means of coded apertures to take two-dimensional compressed measurements. In the CASSI system, the sensing matrix design is established by the coded apertures used in the sensing process. Moreover, the quality of reconstructions depends on the optimal coded aperture designs. In many applications, the information of interest is often found in a specific area of the scene, thus a spatially-selective sensing process is desired. However, the traditional CASSI system generally scans the whole scene even when recovering just an area of interest is required. Hence, this paper presents the analysis of several coded aperture designs that provide a sensing matrix for obtaining more information around a specific area of interest in comparison with the rest of the scene. Several simulations show that the image quality attained with the proposed approach is similar to the traditional non-selective coded apertures, requiring 3 snapshots less. Furthermore, when spatially-selective coded aperture designs are used with a fixed number of snapshots, the reconstructions are improved in up to 1 dB of PSNR (peak signal-to-noise ratio) with respect to the non-selective system in the area of interest.