Learning to denoise astronomical images with U-nets

J 2021

Learning to denoise astronomical images with U-nets

VOJTEKOVÁ, Antónia, Maggie LIEU, Ivan VALTCHANOV, Bruno ALTIERI, Lyndsay OLD et. al.

Základní údaje

Originální název

Learning to denoise astronomical images with U-nets

Autoři

VOJTEKOVÁ, Antónia (703 Slovensko, garant, domácí), Maggie LIEU, Ivan VALTCHANOV, Bruno ALTIERI, Lyndsay OLD, Qifeng CHEN a Filip HROCH (203 Česká republika, domácí)

Vydání

Monthly Notices of the Royal Astronomical Society, Oxford University Press, 2021, 0035-8711

Další údaje

Jazyk

angličtina

Typ výsledku

Článek v odborném periodiku

Obor

10308 Astronomy

Stát vydavatele

Velká Británie a Severní Irsko

Utajení

není předmětem státního či obchodního tajemství

Odkazy

URL

Impakt faktor

Impact factor: 5.235

Kód RIV

RIV/00216224:14310/21:00123743

Organizační jednotka

Přírodovědecká fakulta

DOI

http://dx.doi.org/10.1093/mnras/staa3567

UT WoS

000649000600006

Klíčová slova anglicky

methods: data analysis; techniques: image processing

Štítky

rivok

Příznaky

Mezinárodní význam, Recenzováno

Změněno: 14. 1. 2022 15:57, Mgr. Marie Novosadová Šípková, DiS.

Anotace

V originále

Y Astronomical images are essential for exploring and understanding the Universe. Optical telescopes capable of deep observations, such as the Hubble Space Telescope (HST), are heavily oversubscribed in the Astronomical Community. Images also often contain additive noise, which makes denoising a mandatory step in post-processing the data before further data analysis. In order to maximize the efficiency and information gain in the post-processing of astronomical imaging, we turn to machine learning. We propose ASTRO U-NET, a convolutional neural network for image denoising and enhancement. For a proof-of-concept, we use HST images from Wide Field Camera 3 instrument UV/visible channel with F555W and F606W filters. Our network is able to produce images with noise characteristics as if they are obtained with twice the exposure time, and with minimum bias or information loss. From these images, we are able to recover 95.9 per cent of stars with an average flux error of 2.26 per cent. Furthermore, the images have, on average, 1.63 times higher signal-to-noise ratio than the input noisy images, equivalent to the stacking of at least three input images, which means a significant reduction in the telescope time needed for future astronomical imaging campaigns.

Citovat

VOJTEKOVÁ, Antónia, Maggie LIEU, Ivan VALTCHANOV, Bruno ALTIERI, Lyndsay OLD, Qifeng CHEN a Filip HROCH. Learning to denoise astronomical images with U-nets. Monthly Notices of the Royal Astronomical Society. Oxford University Press, 2021, roč. 503, č. 3, s. 3204-3215. ISSN 0035-8711. Dostupné z: https://dx.doi.org/10.1093/mnras/staa3567.

@article{1821619,
   author = {Vojteková, Antónia and Lieu, Maggie and Valtchanov, Ivan and Altieri, Bruno and Old, Lyndsay and Chen, Qifeng and Hroch, Filip},
   article_number = {3},
   doi = {http://dx.doi.org/10.1093/mnras/staa3567},
   keywords = {methods: data analysis; techniques: image processing},
   language = {eng},
   issn = {0035-8711},
   journal = {Monthly Notices of the Royal Astronomical Society},
   title = {Learning to denoise astronomical images with U-nets},
   url = {https://doi.org/10.1093/mnras/staa3567},
   volume = {503},
   year = {2021}
}

TY  - JOUR
ID  - 1821619
AU  - Vojteková, Antónia - Lieu, Maggie - Valtchanov, Ivan - Altieri, Bruno - Old, Lyndsay - Chen, Qifeng - Hroch, Filip
PY  - 2021
TI  - Learning to denoise astronomical images with U-nets
JF  - Monthly Notices of the Royal Astronomical Society
VL  - 503
IS  - 3
SP  - 3204-3215
EP  - 3204-3215
PB  - Oxford University Press
SN  - 00358711
KW  - methods: data analysis
KW  - techniques: image processing
UR  - https://doi.org/10.1093/mnras/staa3567
N2  - Y Astronomical images are essential for exploring and understanding the Universe. Optical telescopes capable of deep observations, such as the Hubble Space Telescope (HST), are heavily oversubscribed in the Astronomical Community. Images also often contain additive noise, which makes denoising a mandatory step in post-processing the data before further data analysis. In order to maximize the efficiency and information gain in the post-processing of astronomical imaging, we turn to machine learning. We propose ASTRO U-NET, a convolutional neural network for image denoising and enhancement. For a proof-of-concept, we use HST images from Wide Field Camera 3 instrument UV/visible channel with F555W and F606W filters. Our network is able to produce images with noise characteristics as if they are obtained with twice the exposure time, and with minimum bias or information loss. From these images, we are able to recover 95.9 per cent of stars with an average flux error of 2.26 per cent. Furthermore, the images have, on average, 1.63 times higher signal-to-noise ratio than the input noisy images, equivalent to the stacking of at least three input images, which means a significant reduction in the telescope time needed for future astronomical imaging campaigns.
ER  -

VOJTEKOVÁ, Antónia, Maggie LIEU, Ivan VALTCHANOV, Bruno ALTIERI, Lyndsay OLD, Qifeng CHEN a Filip HROCH. Learning to denoise astronomical images with U-nets. \textit{Monthly Notices of the Royal Astronomical Society}. Oxford University Press, 2021, roč.~503, č.~3, s.~3204-3215. ISSN~0035-8711. Dostupné z: https://dx.doi.org/10.1093/mnras/staa3567.

Podrobný výpis o publikaci