Single Image Super-Resolution Based on Improved WGAN

DOI

10.2991/acaai-18.2018.24How to use a DOI?

Keywords

Super-resolution; WGAN-GP; VGG

Abstract

SRGAN has successfully applied the Generative Adversarial Network to the single image super-resolution reconstruction, which has achieved good results. But the loss function based on feature space in SRGAN objectively sacrifices the pursuit of high peak signal-to-noise-ratio (PSNR), which is the result of a tradeoff. At the same time, Improved Training of Wasserstein GANs makes the training process more stable. We redesign the SRGAN, using VGG16 network for feature extraction, setting discriminator network's working space as feature space, and adding the loss function based on the mean square error of pixel space, then gain more details and high PSNR in the reconstruction at the same time. We use the design of WGAN-GP for reference to make the training more stable.

Copyright

© 2018, the Authors. Published by Atlantis Press.

Open Access

This is an open access article distributed under the CC BY-NC license (http://creativecommons.org/licenses/by-nc/4.0/).

Download article (PDF)

TY  - CONF
AU  - Lei Yu
AU  - Xiang Long
AU  - Chao Tong
PY  - 2018/03
DA  - 2018/03
TI  - Single Image Super-Resolution Based on Improved WGAN
BT  - Proceedings of the 2018 International Conference on Advanced Control, Automation and Artificial Intelligence (ACAAI 2018)
PB  - Atlantis Press
SP  - 101
EP  - 104
SN  - 1951-6851
UR  - https://doi.org/10.2991/acaai-18.2018.24
DO  - 10.2991/acaai-18.2018.24
ID  - Yu2018/03
ER  -