Night colorize: Fully convolutional colorization network for low-light images

Lubin Xia; Li Li; Weiqi Jin; Su Qiu; Hongchang Cheng

doi:10.1117/12.2547902

Night colorize: Fully convolutional colorization network for low-light images

Lubin Xia, Li Li, Weiqi Jin, Su Qiu^*, Hongchang Cheng

^*Corresponding author for this work

School of Optics and Photonics

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

1 Citation (Scopus)

Abstract

An end-to-end network is proposed for low-light images natural colorization using a deep fully convolutional architecture. The network consists of a downsampling sub-network and an upsampling sub-network. The downsampling component extracts the high-level features of the input images, while the upsampling component transforms the high-level features to color. A skip connection is used to transmit low layer information to the deep layer so as to improve the colorization accuracy. Gamma correction and random noise augmentation are used to improve the network adaptability to low-light images. The trained model can naturally colorize low-light images without any reference image or artificial scribbles.

Original language	English
Title of host publication	2019 International Conference on Image and Video Processing, and Artificial Intelligence
Editors	Ruidan Su
Publisher	SPIE
ISBN (Electronic)	9781510634091
DOIs	https://doi.org/10.1117/12.2547902
Publication status	Published - 2019
Event	2019 2nd International Conference on Image and Video Processing, and Artificial Intelligence, IVPAI 2019 - Shanghai, China Duration: 23 Aug 2019 → 25 Aug 2019

Publication series

Name	Proceedings of SPIE - The International Society for Optical Engineering
Volume	11321
ISSN (Print)	0277-786X
ISSN (Electronic)	1996-756X

Conference

Conference	2019 2nd International Conference on Image and Video Processing, and Artificial Intelligence, IVPAI 2019
Country/Territory	China
City	Shanghai
Period	23/08/19 → 25/08/19

Keywords

Fully convolutional network
Low-light images
Natural Colorization

Access to Document

10.1117/12.2547902

Cite this

Xia, L., Li, L., Jin, W., Qiu, S., & Cheng, H. (2019). Night colorize: Fully convolutional colorization network for low-light images. In R. Su (Ed.), 2019 International Conference on Image and Video Processing, and Artificial Intelligence Article 113210T (Proceedings of SPIE - The International Society for Optical Engineering; Vol. 11321). SPIE. https://doi.org/10.1117/12.2547902

@inproceedings{fdf976758146484583ca37c1c3d59372,

title = "Night colorize: Fully convolutional colorization network for low-light images",

abstract = "An end-to-end network is proposed for low-light images natural colorization using a deep fully convolutional architecture. The network consists of a downsampling sub-network and an upsampling sub-network. The downsampling component extracts the high-level features of the input images, while the upsampling component transforms the high-level features to color. A skip connection is used to transmit low layer information to the deep layer so as to improve the colorization accuracy. Gamma correction and random noise augmentation are used to improve the network adaptability to low-light images. The trained model can naturally colorize low-light images without any reference image or artificial scribbles.",

keywords = "Fully convolutional network, Low-light images, Natural Colorization",

author = "Lubin Xia and Li Li and Weiqi Jin and Su Qiu and Hongchang Cheng",

note = "Publisher Copyright: {\textcopyright} 2019 SPIE.; 2019 2nd International Conference on Image and Video Processing, and Artificial Intelligence, IVPAI 2019 ; Conference date: 23-08-2019 Through 25-08-2019",

year = "2019",

doi = "10.1117/12.2547902",

language = "English",

series = "Proceedings of SPIE - The International Society for Optical Engineering",

publisher = "SPIE",

editor = "Ruidan Su",

booktitle = "2019 International Conference on Image and Video Processing, and Artificial Intelligence",

address = "United States",

}

Xia, L, Li, L , Jin, W , Qiu, S & Cheng, H 2019, Night colorize: Fully convolutional colorization network for low-light images. in R Su (ed.), 2019 International Conference on Image and Video Processing, and Artificial Intelligence., 113210T, Proceedings of SPIE - The International Society for Optical Engineering, vol. 11321, SPIE, 2019 2nd International Conference on Image and Video Processing, and Artificial Intelligence, IVPAI 2019, Shanghai, China, 23/08/19. https://doi.org/10.1117/12.2547902

Night colorize: Fully convolutional colorization network for low-light images. / Xia, Lubin; Li, Li ; Jin, Weiqi et al.
2019 International Conference on Image and Video Processing, and Artificial Intelligence. ed. / Ruidan Su. SPIE, 2019. 113210T (Proceedings of SPIE - The International Society for Optical Engineering; Vol. 11321).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

TY - GEN

T1 - Night colorize

T2 - 2019 2nd International Conference on Image and Video Processing, and Artificial Intelligence, IVPAI 2019

AU - Xia, Lubin

AU - Li, Li

AU - Jin, Weiqi

AU - Qiu, Su

AU - Cheng, Hongchang

PY - 2019

Y1 - 2019

N2 - An end-to-end network is proposed for low-light images natural colorization using a deep fully convolutional architecture. The network consists of a downsampling sub-network and an upsampling sub-network. The downsampling component extracts the high-level features of the input images, while the upsampling component transforms the high-level features to color. A skip connection is used to transmit low layer information to the deep layer so as to improve the colorization accuracy. Gamma correction and random noise augmentation are used to improve the network adaptability to low-light images. The trained model can naturally colorize low-light images without any reference image or artificial scribbles.

AB - An end-to-end network is proposed for low-light images natural colorization using a deep fully convolutional architecture. The network consists of a downsampling sub-network and an upsampling sub-network. The downsampling component extracts the high-level features of the input images, while the upsampling component transforms the high-level features to color. A skip connection is used to transmit low layer information to the deep layer so as to improve the colorization accuracy. Gamma correction and random noise augmentation are used to improve the network adaptability to low-light images. The trained model can naturally colorize low-light images without any reference image or artificial scribbles.

KW - Fully convolutional network

KW - Low-light images

KW - Natural Colorization

UR - https://www.scopus.com/pages/publications/85077811301

U2 - 10.1117/12.2547902

DO - 10.1117/12.2547902

M3 - Conference contribution

AN - SCOPUS:85077811301

T3 - Proceedings of SPIE - The International Society for Optical Engineering

BT - 2019 International Conference on Image and Video Processing, and Artificial Intelligence

A2 - Su, Ruidan

PB - SPIE

Y2 - 23 August 2019 through 25 August 2019

ER -

Night colorize: Fully convolutional colorization network for low-light images

Abstract

Publication series

Conference

Keywords

Access to Document

Other files and links

Fingerprint

Cite this