Two stage real facial image super resolution with geometric knowledge infusion

Shangling Jiang; Jiafeng Hao; Jiquan Ma; Xuliang Li

doi:10.1117/12.2640272

1 August 2022 Two stage real facial image super resolution with geometric knowledge infusion

Shangling Jiang, Jiafeng Hao, Jiquan Ma, Xuliang Li

Proceedings Volume 12257, 4th International Conference on Information Science, Electrical, and Automation Engineering (ISEAE 2022); 122571W (2022) https://doi.org/10.1117/12.2640272
Event: 4th International Conference on Information Science, Electrical, and Automation Engineering (ISEAE 2022), 2022, Guangzhou, China

Abstract

Super-resolution(SR) is a classic issue in computer vision. Existing SR methods mainly rely on artificially downsampling to prepare the pair-wised training data set without considering the blur kernel inconsistency and often result in an unstable performance in the real scenario. Especially for facial images, unacceptable reconstruction seriously affects its downstream applications. In order to improve the quality in real facial image SR, we proposed an end-to-end deep neural network to mimic a realistic process of image degradation and reconstruct a high-resolution (HR) facial image from a low-resolution (LR) one. At the first stage, a generative adversarial network (GAN) is employed to sample more realistic LR images by jointly considering facial landmark and composite blur kernels. At the second stage, pairwised training samples are fed into a decoder to fit the mapping from LR to HR. Experiments were performed on real LR test set from Widerface and LS3D-W dataset. Results demonstrate that our model outperforms existing competing methods. Furthermore, the ablation studies valid the effectiveness of the proposed components.

Citation Download Citation

Shangling Jiang, Jiafeng Hao, Jiquan Ma, and Xuliang Li "Two stage real facial image super resolution with geometric knowledge infusion", Proc. SPIE 12257, 4th International Conference on Information Science, Electrical, and Automation Engineering (ISEAE 2022), 122571W (1 August 2022); https://doi.org/10.1117/12.2640272

ACCESS THE FULL ARTICLE

INSTITUTIONAL
Select your institution to access the SPIE Digital Library.

SELECT YOUR INSTITUTION

PERSONAL
Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.

PERSONAL SIGN IN

No SPIE Account? Create one

PURCHASE THIS CONTENT

SUBSCRIBE TO DIGITAL LIBRARY

50 downloads per 1-year subscription

Members: $195

Non-members: $335 ADD TO CART

25 downloads per 1 - year subscription

Members: $145

Non-members: $250 ADD TO CART

PURCHASE SINGLE ARTICLE

Includes PDF, HTML & Video, when available

Members: $17.00

Non-members: $21.00 ADD TO CART

PROCEEDINGS
9 PAGES

DOWNLOAD PAPER SAVE TO MY LIBRARY

GET CITATION

RIGHTS & PERMISSIONS

Get copyright permission Get copyright permission on Copyright Marketplace

KEYWORDS

Lawrencium

Gallium nitride

Image quality

Image processing

Super resolution

Principal component analysis

Visualization

Show All Keywords

Keywords/Phrases

Search In:

Publication Years