Video super-resolution enhancement based on 3D spatio-temporal fusion model

Wen Xu; Zhaohui Meng

doi:10.1117/12.2635686

6 May 2022 Video super-resolution enhancement based on 3D spatio-temporal fusion model

Wen Xu, Zhaohui Meng

Proceedings Volume 12256, International Conference on Electronic Information Engineering, Big Data, and Computer Technology (EIBDCT 2022); 122562J (2022) https://doi.org/10.1117/12.2635686
Event: 2022 International Conference on Electronic Information Engineering, Big Data and Computer Technology, 2022, Sanya, China

Abstract

In the field of video super-resolution, multiple frames provide more scene information, not only in intra-frame spatial dependence, but also inter-frame temporal dependence (for example motion, brightness, and color changes). Therefore, existing work mainly focuses on making better use of spatio-temporal dependence, including explicit motion compensation (for example based on optical flow, based on learning) and cyclic methods. After years of research by predecessors, video super-resolution has achieved good results, but there are still some difficulties and shortcomings that need to be resolved. Generally speaking, SR algorithm series using deep learning technology are different from each other in the following main aspects: different types of network architectures, different types of loss functions, and different types of learning strategies. In this paper, the spatio-temporal fusion module is improved on the basis of a two-stage 3D convolutional network. The adjacent frame and the current frame are merged in space and time, and the generated residual block is added to the feature map. The two-stage 3D convolution is applied to the residual block in the backbone network DenseNet to better process the pre-deblurring module.

Citation Download Citation

Wen Xu and Zhaohui Meng "Video super-resolution enhancement based on 3D spatio-temporal fusion model", Proc. SPIE 12256, International Conference on Electronic Information Engineering, Big Data, and Computer Technology (EIBDCT 2022), 122562J (6 May 2022); https://doi.org/10.1117/12.2635686

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
6 PAGES

DOWNLOAD PAPER SAVE TO MY LIBRARY

GET CITATION

RIGHTS & PERMISSIONS

Get copyright permission Get copyright permission on Copyright Marketplace

KEYWORDS

Convolution

Video

3D modeling

Super resolution

Silicon

Feature extraction

Motion models

Show All Keywords

Keywords/Phrases

Search In:

Publication Years