Coded aperture spectral imaging is a new system to captures multiframes images and reconstructs them into spectral image cube based on compressive sensing theory (CS). However, using dynamic transformed coded aperture pattern can cause two primary problems, firstly the whole exposure procedure needs to be staring on the same surface feature which is depended on a high quality stable platform; secondly the coded aperture’s transformation might reduce the system’s stability. To avoid these problems without a loss of information for precise reconstruction, in the paper we propose dividing the single image panel into encoding spatial overlapped sub-districts. We design a pushbroom scan pattern to ensure each sub-district have enough sampling measurements. In each sub-district, we infer its measurement matrix can satisfy the sparsely requirements needed for accurate estimation and final reconstruction with CS sampling. Considering with efficiency and accuracy, we design a orthogonal self-loop coding mask (lines irrelevant) to guarantee the coding are irrelevant among distinct snapshot of the same scene. The simulation experiment reveals the design helping reconstruct the scene spectral cube with high throughput and resolution.

Date Published: 26 September 2013
PDF: 9 pages
Proc. SPIE 8855, Optics and Photonics for Information Processing VII, 88550I (26 September 2013); doi: 10.1117/12.2021596

Published in SPIE Proceedings Vol. 8855:
Optics and Photonics for Information Processing VII
Khan M. Iftekharuddin; Abdul A. S. Awwal; Andrés Márquez, Editor(s)