Generative Adversarial Networks for Spectral Super-Resolution and Bidirectional RGB-to-Multispectral Mapping
Abstract
Acquisition of multi-and hyperspectral imagery imposes significant requirements on the hardware capabilities of the sensors involved. In order to keep costs manageable, and due to limitations in the sensing technology, tradeoffs between the spectral and the spatial resolution of hyperspectral images are usually made. Such tradeoffs are usually not necessary when considering acquisition of traditional RGB imagery. We investigate the use of statistical learning, and in particular, of conditional generative adversarial networks (cGANs) to estimate mappings from three-channel RGB to 31-band multispectral imagery. We demonstrate the application of the proposed approach to (i) RGB-to-multispectral image mapping, (ii) spectral super-resolution of image data, and (iii) recovery of RGB imagery from multispectral data.
Cite
Text
Lore et al. "Generative Adversarial Networks for Spectral Super-Resolution and Bidirectional RGB-to-Multispectral Mapping." IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, 2019. doi:10.1109/CVPRW.2019.00122Markdown
[Lore et al. "Generative Adversarial Networks for Spectral Super-Resolution and Bidirectional RGB-to-Multispectral Mapping." IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, 2019.](https://mlanthology.org/cvprw/2019/lore2019cvprw-generative/) doi:10.1109/CVPRW.2019.00122BibTeX
@inproceedings{lore2019cvprw-generative,
title = {{Generative Adversarial Networks for Spectral Super-Resolution and Bidirectional RGB-to-Multispectral Mapping}},
author = {Lore, Kin Gwn and Reddy, Kishore K. and Giering, Michael and Bernal, Edgar A.},
booktitle = {IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops},
year = {2019},
pages = {926-933},
doi = {10.1109/CVPRW.2019.00122},
url = {https://mlanthology.org/cvprw/2019/lore2019cvprw-generative/}
}