MFEN: Lightweight multi-scale feature extraction super-resolution network in embedded system

Deep convolutional neural networks (CNN) have achieved remarkable performance in super-resolution (SR) recently. However, deep CNN-based methods are difficult to be utilized in embedded portable device due to their heavy computation and memory consumption. To solve the above problem, we propose an e...

Full description

Saved in:
Bibliographic Details
Published in:Microprocessors and microsystems 2022-09, Vol.93, p.104568, Article 104568
Main Authors: Xiao, Hang, Qin, Jiayi, Jeon, Seunggil, Yan, Binyu, Yang, Xiaomin
Format: Article
Language:English
Subjects:
Deep learning
Feature fusion
Multi-scale feature
Super-resolution
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Deep convolutional neural networks (CNN) have achieved remarkable performance in super-resolution (SR) recently. However, deep CNN-based methods are difficult to be utilized in embedded portable device due to their heavy computation and memory consumption. To solve the above problem, we propose an effective lightweight multi-scale feature extraction super-resolution network (MFEN) by constructing multi-scale feature extraction blocks (MFEB), which progressively obtains multi-scale and hierarchical information. In addition, we also propose an efficient progressive feature fusion (PFF) strategy to aggregate multi-scale informative features. Qualitative and quantitative evaluation results on the benchmark datasets show that our designed method can obtain better performance than most state-of-the-art methods. Moreover, the computation complexity and running time of the MFEN are significantly reduced to provide convenience in real-time image processing technology for embedded devices.
ISSN:0141-9331
1872-9436
DOI:10.1016/j.micpro.2022.104568