Light Field Synthesis from a Single Image using Improved Wasserstein Generative Adversarial Network

Lingyan Ruan; Bin Chen; Miu-Ling Lam

Light Field Synthesis from a Single Image using Improved Wasserstein Generative Adversarial Network

School of Creative Media

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

6 Citations (Scopus)

Abstract

We present a deep learning-based method to synthesize a 4D light field from a single 2D RGB image. We consider the light field synthesis problem equivalent to image super-resolution, and solve it by using the improved Wasserstein Generative Adversarial Network with gradient penalty (WGAN-GP). Experimental results demonstrate that our algorithm can predict complex occlusions and relative depths in challenging scenes. The light fields synthesized by our method has much higher signal-to-noise ratio and structural similarity than the state-of-the-art approach.

Original language	English
Title of host publication	EUROGRAPHICS 2018
Subtitle of host publication	Delft, The Netherlands
Publisher	Eurographics Association
Pages	19-20
ISBN (Electronic)	1017-4656
Publication status	Published - Apr 2018
Event	39th Eurographics Conference (EG 2018) - Delft University of Technology, Delft, Netherlands Duration: 16 Apr 2018 → 20 Apr 2018 https://www.eurographics2018.nl/

Conference

Conference	39th Eurographics Conference (EG 2018)
Place	Netherlands
City	Delft
Period	16/04/18 → 20/04/18
Internet address	https://www.eurographics2018.nl/

Bibliographical note

Research Unit(s) information for this publication is provided by the author(s) concerned.

Other Access Options

Check CityUHK Library

Permanent Link

Right click to copy link

Cite this

@inproceedings{26c0144bcc9d422ba7cb47adbc251982,

title = "Light Field Synthesis from a Single Image using Improved Wasserstein Generative Adversarial Network",

abstract = "We present a deep learning-based method to synthesize a 4D light field from a single 2D RGB image. We consider the light field synthesis problem equivalent to image super-resolution, and solve it by using the improved Wasserstein Generative Adversarial Network with gradient penalty (WGAN-GP). Experimental results demonstrate that our algorithm can predict complex occlusions and relative depths in challenging scenes. The light fields synthesized by our method has much higher signal-to-noise ratio and structural similarity than the state-of-the-art approach.",

author = "Lingyan Ruan and Bin Chen and Miu-Ling Lam",

note = "Research Unit(s) information for this publication is provided by the author(s) concerned.; 39th Eurographics Conference (EG 2018) ; Conference date: 16-04-2018 Through 20-04-2018",

year = "2018",

month = apr,

language = "English",

pages = "19--20",

booktitle = "EUROGRAPHICS 2018",

publisher = "Eurographics Association",

url = "https://www.eurographics2018.nl/",

}

Light Field Synthesis from a Single Image using Improved Wasserstein Generative Adversarial Network. / Ruan, Lingyan ; Chen, Bin ; Lam, Miu-Ling.
EUROGRAPHICS 2018: Delft, The Netherlands. Eurographics Association, 2018. p. 19-20.

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

TY - GEN

T1 - Light Field Synthesis from a Single Image using Improved Wasserstein Generative Adversarial Network

AU - Ruan, Lingyan

AU - Chen, Bin

AU - Lam, Miu-Ling

N1 - Research Unit(s) information for this publication is provided by the author(s) concerned.

PY - 2018/4

Y1 - 2018/4

N2 - We present a deep learning-based method to synthesize a 4D light field from a single 2D RGB image. We consider the light field synthesis problem equivalent to image super-resolution, and solve it by using the improved Wasserstein Generative Adversarial Network with gradient penalty (WGAN-GP). Experimental results demonstrate that our algorithm can predict complex occlusions and relative depths in challenging scenes. The light fields synthesized by our method has much higher signal-to-noise ratio and structural similarity than the state-of-the-art approach.

AB - We present a deep learning-based method to synthesize a 4D light field from a single 2D RGB image. We consider the light field synthesis problem equivalent to image super-resolution, and solve it by using the improved Wasserstein Generative Adversarial Network with gradient penalty (WGAN-GP). Experimental results demonstrate that our algorithm can predict complex occlusions and relative depths in challenging scenes. The light fields synthesized by our method has much higher signal-to-noise ratio and structural similarity than the state-of-the-art approach.

UR - https://www.scopus.com/pages/publications/85078263765

UR - https://www.scopus.com/record/pubmetrics.uri?eid=2-s2.0-85078263765&origin=recordpage

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

SP - 19

EP - 20

BT - EUROGRAPHICS 2018

PB - Eurographics Association

T2 - 39th Eurographics Conference (EG 2018)

Y2 - 16 April 2018 through 20 April 2018

ER -

Light Field Synthesis from a Single Image using Improved Wasserstein Generative Adversarial Network

Abstract

Conference

Bibliographical note

Other Access Options

Permanent Link

Other Files and Links

Fingerprint

Cite this