Low-light image enhancement of permanently shadowed lunar regions with physics-based machine learning

Moseley, Ben; Bickel, Valentin; Lopez-Francos, Ignacio; Rana, Loveneesh; Olivares Mendez, Miguel Angel; Wingo, Dennis; Zuniga, Allison; Subtil, Nuno; D’Eon, Eugene

Reference : Low-light image enhancement of permanently shadowed lunar regions with physics-based ...


	Document type :	Scientific congresses, symposiums and conference proceedings : Paper published in a book
	Discipline(s) :	Engineering, computing & technology : Computer science
	To cite this reference:	http://hdl.handle.net/10993/45541

Title :	Low-light image enhancement of permanently shadowed lunar regions with physics-based machine learning
Language :	English
Author, co-author :	Moseley, Ben [University of Oxford]
	Bickel, Valentin [MPS Goettingen & ETH Zurich]
	Lopez-Francos, Ignacio [NASA Ames Research Center]
	Rana, Loveneesh [University of Luxembourg > Interdisciplinary Centre for Security, Reliability and Trust (SNT) > Remote Sensing]
	Olivares Mendez, Miguel Angel [University of Luxembourg > Interdisciplinary Centre for Security, Reliability and Trust (SNT) > Space Robotics >]
	Wingo, Dennis [Skycorp Inc.]
	Zuniga, Allison [NASA Ames Research Center]
	Subtil, Nuno [Nvidia]
	D’Eon, Eugene [Nvidia]
Publication date :	Dec-2020
Main document title :	Low-light image enhancement of permanently shadowed lunar regions with physics-based machine learning
Peer reviewed :	Yes
Event name :	Conference on Neural Information Processing Systems, NeurIPS
Event date :	from 6-12-2020 to 12-12-2020
Event place (city) :	remote
Keywords :	[en] Machine learning ; Remote sensing ; Lunar exploration
Abstract :	[en] Finding water(-ice) on the Moon is key to enabling a sustainable human presence on the Moon and beyond. There is evidence that water-ice is abundant in and around the Moon’s Permanently Shadowed Regions (PSRs), however, direct visual detection has not yet been possible. Surface ice or related physical features could potentially be directly detected from high-resolution optical imagery, but, due to the extremely low-light conditions in these areas, high levels of sensor and photon noise make this very challenging. In this work we generate high-resolution, low-noise optical images over lunar PSRs by using two physics-based deep neural networks to model and remove CCD-related and photon noise in existing low-light optical imagery, potentially paving the way for a direct water-ice detection method.
Permalink :	http://hdl.handle.net/10993/45541

File(s) associated to this reference

Fulltext file(s):

	File	Commentary	Version	Size	Access
Open access	m4g2_neurips.pdf		Author preprint	4.08 MB	View/Open

All documents in ORBi^lu are protected by a user license.

University of Luxembourg Library | Feedback | Legal notices