DRLFNet: A Dense-Connection Residual Learning Neural Network for Light Field Super Resolution

Congrui Fu; Xin Ma; Yao Liu; Junhui Hou; Hui Yuan

doi:10.1007/978-3-030-87361-5_41

DRLFNet: A Dense-Connection Residual Learning Neural Network for Light Field Super Resolution

Congrui Fu
, Xin Ma
, Yao Liu
, Junhui Hou
, Hui Yuan^*

^*Corresponding author for this work

Department of Computer Science

Research output: Chapters, Conference Papers, Creative and Literary Works › RGC 32 - Refereed conference paper (with host publication) › peer-review

Abstract

Light field records both spatial and angular information of light rays. By using light field cameras, 3D scenes can be reconstructed easily for further virtual reality applications. Limited by the sensor size, there is a trade-off between the spatial and angular resolution. To address this problem, we propose a dense-connection residual learning neural network, namely DRLFNet, to super resolve light field images in spatial domain. The dense-connection residual learning is implemented based on the proposed dense-connection residual block (DResBlock) that is used to efficiently exploit the joint spatial and angular features and the hierarchical features in different layers. Experimental results demonstrate that the proposed method out-performs other state-of-the-art methods by a large margin in both visual and numerical evaluations.

Original language	English
Title of host publication	Image and Graphics
Subtitle of host publication	11th International Conference, ICIG 2021, Haikou, China, August 6–8, 2021, Proceedings, Part III
Editors	Yuxin Peng, Shi-Min Hu, Moncef Gabbouj, Kun Zhou, Michael Elad, Kun Xu
Place of Publication	Cham
Publisher	Springer
Pages	501-510
Volume	Part III
ISBN (Electronic)	9783030873615
ISBN (Print)	9783030873608
DOIs	https://doi.org/10.1007/978-3-030-87361-5_41
Publication status	Published - 2021
Event	11th International Conference on Image and Graphics (ICIG 2021) - Haikou Eadry Royal Garden Hotel, Haikou, China Duration: 6 Aug 2021 → 8 Aug 2021 http://icig2021.csig.org.cn/

Publication series

Name	Lecture Notes in Computer Science
Volume	12890
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Conference

Conference	11th International Conference on Image and Graphics (ICIG 2021)
Place	China
City	Haikou
Period	6/08/21 → 8/08/21
Internet address	http://icig2021.csig.org.cn/

Bibliographical note

Full text of this publication does not contain sufficient affiliation information. With consent from the author(s) concerned, the Research Unit(s) information for this record is based on the existing academic department affiliation of the author(s).

Research Keywords

Dense-connection
Light field images
Residual learning
Super-resolution

Access Output

10.1007/978-3-030-87361-5_41

Other Access Options

Check CityUHK Library

Permanent Link

Right click to copy link

Cite this

Fu, C., Ma, X., Liu, Y., Hou, J., & Yuan, H. (2021). DRLFNet: A Dense-Connection Residual Learning Neural Network for Light Field Super Resolution. In Y. Peng, S.-M. Hu, M. Gabbouj, K. Zhou, M. Elad, & K. Xu (Eds.), Image and Graphics: 11th International Conference, ICIG 2021, Haikou, China, August 6–8, 2021, Proceedings, Part III (Vol. Part III, pp. 501-510). (Lecture Notes in Computer Science; Vol. 12890). Springer . https://doi.org/10.1007/978-3-030-87361-5_41

Fu, Congrui ; Ma, Xin ; Liu, Yao et al. / DRLFNet : A Dense-Connection Residual Learning Neural Network for Light Field Super Resolution. Image and Graphics: 11th International Conference, ICIG 2021, Haikou, China, August 6–8, 2021, Proceedings, Part III. editor / Yuxin Peng ; Shi-Min Hu ; Moncef Gabbouj ; Kun Zhou ; Michael Elad ; Kun Xu. Vol. Part III Cham : Springer , 2021. pp. 501-510 (Lecture Notes in Computer Science).

@inproceedings{73344c09a2ba44b4b49f4214129bb50f,

title = "DRLFNet: A Dense-Connection Residual Learning Neural Network for Light Field Super Resolution",

abstract = "Light field records both spatial and angular information of light rays. By using light field cameras, 3D scenes can be reconstructed easily for further virtual reality applications. Limited by the sensor size, there is a trade-off between the spatial and angular resolution. To address this problem, we propose a dense-connection residual learning neural network, namely DRLFNet, to super resolve light field images in spatial domain. The dense-connection residual learning is implemented based on the proposed dense-connection residual block (DResBlock) that is used to efficiently exploit the joint spatial and angular features and the hierarchical features in different layers. Experimental results demonstrate that the proposed method out-performs other state-of-the-art methods by a large margin in both visual and numerical evaluations.",

keywords = "Dense-connection, Light field images, Residual learning, Super-resolution",

author = "Congrui Fu and Xin Ma and Yao Liu and Junhui Hou and Hui Yuan",

note = "Full text of this publication does not contain sufficient affiliation information. With consent from the author(s) concerned, the Research Unit(s) information for this record is based on the existing academic department affiliation of the author(s).; 11th International Conference on Image and Graphics (ICIG 2021) ; Conference date: 06-08-2021 Through 08-08-2021",

year = "2021",

doi = "10.1007/978-3-030-87361-5\_41",

language = "English",

isbn = "9783030873608",

volume = "Part III",

series = "Lecture Notes in Computer Science",

publisher = "Springer ",

pages = "501--510",

editor = "Yuxin Peng and Shi-Min Hu and Moncef Gabbouj and Kun Zhou and Michael Elad and Kun Xu",

booktitle = "Image and Graphics",

url = "http://icig2021.csig.org.cn/",

}

Fu, C, Ma, X, Liu, Y, Hou, J & Yuan, H 2021, DRLFNet: A Dense-Connection Residual Learning Neural Network for Light Field Super Resolution. in Y Peng, S-M Hu, M Gabbouj, K Zhou, M Elad & K Xu (eds), Image and Graphics: 11th International Conference, ICIG 2021, Haikou, China, August 6–8, 2021, Proceedings, Part III. vol. Part III, Lecture Notes in Computer Science, vol. 12890, Springer , Cham, pp. 501-510, 11th International Conference on Image and Graphics (ICIG 2021), Haikou, China, 6/08/21. https://doi.org/10.1007/978-3-030-87361-5_41

DRLFNet: A Dense-Connection Residual Learning Neural Network for Light Field Super Resolution. / Fu, Congrui; Ma, Xin; Liu, Yao et al.
Image and Graphics: 11th International Conference, ICIG 2021, Haikou, China, August 6–8, 2021, Proceedings, Part III. ed. / Yuxin Peng; Shi-Min Hu; Moncef Gabbouj; Kun Zhou; Michael Elad; Kun Xu. Vol. Part III Cham: Springer , 2021. p. 501-510 (Lecture Notes in Computer Science; Vol. 12890).

Research output: Chapters, Conference Papers, Creative and Literary Works › RGC 32 - Refereed conference paper (with host publication) › peer-review

TY - GEN

T1 - DRLFNet

T2 - 11th International Conference on Image and Graphics (ICIG 2021)

AU - Fu, Congrui

AU - Ma, Xin

AU - Liu, Yao

AU - Hou, Junhui

AU - Yuan, Hui

N1 - Full text of this publication does not contain sufficient affiliation information. With consent from the author(s) concerned, the Research Unit(s) information for this record is based on the existing academic department affiliation of the author(s).

PY - 2021

Y1 - 2021

N2 - Light field records both spatial and angular information of light rays. By using light field cameras, 3D scenes can be reconstructed easily for further virtual reality applications. Limited by the sensor size, there is a trade-off between the spatial and angular resolution. To address this problem, we propose a dense-connection residual learning neural network, namely DRLFNet, to super resolve light field images in spatial domain. The dense-connection residual learning is implemented based on the proposed dense-connection residual block (DResBlock) that is used to efficiently exploit the joint spatial and angular features and the hierarchical features in different layers. Experimental results demonstrate that the proposed method out-performs other state-of-the-art methods by a large margin in both visual and numerical evaluations.

AB - Light field records both spatial and angular information of light rays. By using light field cameras, 3D scenes can be reconstructed easily for further virtual reality applications. Limited by the sensor size, there is a trade-off between the spatial and angular resolution. To address this problem, we propose a dense-connection residual learning neural network, namely DRLFNet, to super resolve light field images in spatial domain. The dense-connection residual learning is implemented based on the proposed dense-connection residual block (DResBlock) that is used to efficiently exploit the joint spatial and angular features and the hierarchical features in different layers. Experimental results demonstrate that the proposed method out-performs other state-of-the-art methods by a large margin in both visual and numerical evaluations.

KW - Dense-connection

KW - Light field images

KW - Residual learning

KW - Super-resolution

UR - https://www.scopus.com/pages/publications/85117114232

UR - https://www.scopus.com/record/pubmetrics.uri?eid=2-s2.0-85117114232&origin=recordpage

U2 - 10.1007/978-3-030-87361-5_41

DO - 10.1007/978-3-030-87361-5_41

M3 - RGC 32 - Refereed conference paper (with host publication)

SN - 9783030873608

VL - Part III

T3 - Lecture Notes in Computer Science

SP - 501

EP - 510

BT - Image and Graphics

A2 - Peng, Yuxin

A2 - Hu, Shi-Min

A2 - Gabbouj, Moncef

A2 - Zhou, Kun

A2 - Elad, Michael

A2 - Xu, Kun

PB - Springer

CY - Cham

Y2 - 6 August 2021 through 8 August 2021

ER -

Fu C, Ma X, Liu Y, Hou J, Yuan H. DRLFNet: A Dense-Connection Residual Learning Neural Network for Light Field Super Resolution. In Peng Y, Hu SM, Gabbouj M, Zhou K, Elad M, Xu K, editors, Image and Graphics: 11th International Conference, ICIG 2021, Haikou, China, August 6–8, 2021, Proceedings, Part III. Vol. Part III. Cham: Springer . 2021. p. 501-510. (Lecture Notes in Computer Science). Epub 2021 Sept 30. doi: 10.1007/978-3-030-87361-5_41

DRLFNet: A Dense-Connection Residual Learning Neural Network for Light Field Super Resolution

Abstract

Publication series

Conference

Bibliographical note

Research Keywords

Access Output

Other Access Options

Permanent Link

Other Files and Links

Fingerprint

Cite this