CNNs in the frequency domain for image super-resolution

Yingnan Liu; Randy Clinton Paffenroth

doi:10.1117/12.2539288

27 November 2019 CNNs in the frequency domain for image super-resolution

Proceedings Volume 11321, 2019 International Conference on Image and Video Processing, and Artificial Intelligence; 113210D (2019) https://doi.org/10.1117/12.2539288
Event: The Second International Conference on Image, Video Processing and Artifical Intelligence, 2019, Shanghai, China

Abstract

This paper develops methods for recovering high-resolution images from low-resolution images by combining ideas inspired by sparse coding, such as compressive sensing techniques, with super-resolution neural networks. Sparse coding leverages the existence of bases in which signals can be sparsely represented, and herein we use such ideas to improve the performance of super-resolution convolutional neural networks (CNN). In particular, we propose an improved model in which CNNs are used for super-resolution in the frequency domain, and we demonstrate that such an approach improves the performance of image super-resolution neural networks. In addition, we indicate that instead of numerous deep layers, a shallower architecture in the frequency domain is sufficient for many types of image super-resolution problems.

Citation Download Citation

Yingnan Liu and Randy Clinton Paffenroth "CNNs in the frequency domain for image super-resolution", Proc. SPIE 11321, 2019 International Conference on Image and Video Processing, and Artificial Intelligence, 113210D (27 November 2019); https://doi.org/10.1117/12.2539288

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