Compact and robust hyperspectral camera based on compressed sensing
| Autor: | Jan Václavík, Karel Žídek, J. Hlubuček, Ondřej Denk |
|---|---|
| Rok vydání: | 2016 |
| Předmět: |
medicine.medical_specialty
Pixel Computer science business.industry ComputingMethodologies_IMAGEPROCESSINGANDCOMPUTERVISION Hyperspectral imaging 020206 networking & telecommunications 02 engineering and technology Spectral imaging Compressed sensing Camera auto-calibration Full spectral imaging 0202 electrical engineering electronic engineering information engineering medicine 020201 artificial intelligence & image processing Computer vision Coded aperture Artificial intelligence business Camera resectioning |
| Zdroj: | SPIE Proceedings. |
| ISSN: | 0277-786X |
| DOI: | 10.1117/12.2250268 |
| Popis: | Spectrum of light which is emitted or reflected by an object carries immense amount of information about the object. A simple piece of evidence is the importance of color sensing for human vision. Combining an image acquisition with efficient measurement of light spectra for each detected pixel is therefore one of the important issues in imaging, referred as hyperspectral imaging. We demonstrate a construction of a compact and robust hyperspectral camera for the visible and near-IR spectral region. The camera was designed vastly based on off-shelf optics, yet an extensive optimization and addition of three customized parts enabled construction of the camera featuring a low f-number (F/3.9) and fully concentric optics. We employ a novel approach of compressed sensing (namely coded aperture snapshot spectral imaging, abbrev. CASSI). The compressed sensing enables to computationally extract an encoded hyperspectral information from a single camera exposition. Owing to the technique the camera lacks any moving or scanning part, while it can record the full image and spectral information in a single snapshot. Moreover, unlike the commonly used compressed sensing table-top apparatuses, the camera represents a portable device able to work outside a lab. We demonstrate the spectro-temporal reconstruction of recorded scenes based on 90×90 random matrix encoding. Finally, we discuss potential of the compressed sensing in hyperspectral camera. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|