TCSR: Lightweight Transformer and CNN Interaction Network for Image Super-Resolution

Convolutional neural network (CNN) has achieved impressive success in lightweight image super-resolution (SR) methods, yet the nature of its local operations constrains the SR performance. Recent Transformer has attracted increasing attention in lightweight SR methods owing to its remarkable global...

Full description

Saved in:
Bibliographic Details
Main Authors: Danlin Cai, Wenwen Tan, Feiyang Chen, Xinchi Lou, Jianbin Xiahou, Daxin Zhu, Detian Huang
Format: Article
Language:English
Published: IEEE 2024-01-01
Series:IEEE Access
Subjects:
Online Access:https://ieeexplore.ieee.org/document/10707600/
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1850266429893378048
author Danlin Cai
Wenwen Tan
Feiyang Chen
Xinchi Lou
Jianbin Xiahou
Daxin Zhu
Detian Huang
author_facet Danlin Cai
Wenwen Tan
Feiyang Chen
Xinchi Lou
Jianbin Xiahou
Daxin Zhu
Detian Huang
author_sort Danlin Cai
collection DOAJ
description Convolutional neural network (CNN) has achieved impressive success in lightweight image super-resolution (SR) methods, yet the nature of its local operations constrains the SR performance. Recent Transformer has attracted increasing attention in lightweight SR methods owing to its remarkable global feature extraction capacity. However, the huge computational cost makes it challenging for lightweight SR methods to efficiently utilize Transformer to exploit global contextual information from shallow to intermediate layers. To address these issues, we propose a novel lightweight Transformer and CNN interaction network for image Super-Resolution (TCSR), which fully leverages the complementary strengths of Transformer and CNN. Specifically, an efficient lightweight Transformer and CNN Interaction Block (TCIB) is designed to extract local and global features at various stages of the network, resulting in favorable hybrid features that significantly improve the quality of reconstructed images. Then, we construct a lightweight Reversed UNet (RUNet) to progressively aggregate hybrid features as well as to better trade-off the reconstruction accuracy and efficiency. Furthermore, we introduce a Refinement module to further refine edge and texture details with global information. Experimental results on numerous benchmarks validate that the proposed TCSR achieves superior performance with fewer parameters and less computational overhead than state-of-the-art lightweight methods.
format Article
id doaj-art-4380ee689eae43a0804e6dffcd6a168a
institution OA Journals
issn 2169-3536
language English
publishDate 2024-01-01
publisher IEEE
record_format Article
series IEEE Access
spelling doaj-art-4380ee689eae43a0804e6dffcd6a168a2025-08-20T01:54:11ZengIEEEIEEE Access2169-35362024-01-011217478217479510.1109/ACCESS.2024.347636910707600TCSR: Lightweight Transformer and CNN Interaction Network for Image Super-ResolutionDanlin Cai0https://orcid.org/0000-0003-3083-5503Wenwen Tan1Feiyang Chen2Xinchi Lou3Jianbin Xiahou4Daxin Zhu5https://orcid.org/0000-0002-8060-7368Detian Huang6https://orcid.org/0000-0002-8542-3728School of Mathematics and Computer Science, Quanzhou Normal University, Quanzhou, ChinaCollege of Engineering, Huaqiao University, Quanzhou, ChinaCollege of Engineering, Huaqiao University, Quanzhou, ChinaSchool of Mathematics and Computer Science, Quanzhou Normal University, Quanzhou, ChinaSchool of Mathematics and Computer Science, Quanzhou Normal University, Quanzhou, ChinaSchool of Mathematics and Computer Science, Quanzhou Normal University, Quanzhou, ChinaCollege of Engineering, Huaqiao University, Quanzhou, ChinaConvolutional neural network (CNN) has achieved impressive success in lightweight image super-resolution (SR) methods, yet the nature of its local operations constrains the SR performance. Recent Transformer has attracted increasing attention in lightweight SR methods owing to its remarkable global feature extraction capacity. However, the huge computational cost makes it challenging for lightweight SR methods to efficiently utilize Transformer to exploit global contextual information from shallow to intermediate layers. To address these issues, we propose a novel lightweight Transformer and CNN interaction network for image Super-Resolution (TCSR), which fully leverages the complementary strengths of Transformer and CNN. Specifically, an efficient lightweight Transformer and CNN Interaction Block (TCIB) is designed to extract local and global features at various stages of the network, resulting in favorable hybrid features that significantly improve the quality of reconstructed images. Then, we construct a lightweight Reversed UNet (RUNet) to progressively aggregate hybrid features as well as to better trade-off the reconstruction accuracy and efficiency. Furthermore, we introduce a Refinement module to further refine edge and texture details with global information. Experimental results on numerous benchmarks validate that the proposed TCSR achieves superior performance with fewer parameters and less computational overhead than state-of-the-art lightweight methods.https://ieeexplore.ieee.org/document/10707600/Lightweight image super-resolutiontransformerconvolutional neural networkUNet
spellingShingle Danlin Cai
Wenwen Tan
Feiyang Chen
Xinchi Lou
Jianbin Xiahou
Daxin Zhu
Detian Huang
TCSR: Lightweight Transformer and CNN Interaction Network for Image Super-Resolution
IEEE Access
Lightweight image super-resolution
transformer
convolutional neural network
UNet
title TCSR: Lightweight Transformer and CNN Interaction Network for Image Super-Resolution
title_full TCSR: Lightweight Transformer and CNN Interaction Network for Image Super-Resolution
title_fullStr TCSR: Lightweight Transformer and CNN Interaction Network for Image Super-Resolution
title_full_unstemmed TCSR: Lightweight Transformer and CNN Interaction Network for Image Super-Resolution
title_short TCSR: Lightweight Transformer and CNN Interaction Network for Image Super-Resolution
title_sort tcsr lightweight transformer and cnn interaction network for image super resolution
topic Lightweight image super-resolution
transformer
convolutional neural network
UNet
url https://ieeexplore.ieee.org/document/10707600/
work_keys_str_mv AT danlincai tcsrlightweighttransformerandcnninteractionnetworkforimagesuperresolution
AT wenwentan tcsrlightweighttransformerandcnninteractionnetworkforimagesuperresolution
AT feiyangchen tcsrlightweighttransformerandcnninteractionnetworkforimagesuperresolution
AT xinchilou tcsrlightweighttransformerandcnninteractionnetworkforimagesuperresolution
AT jianbinxiahou tcsrlightweighttransformerandcnninteractionnetworkforimagesuperresolution
AT daxinzhu tcsrlightweighttransformerandcnninteractionnetworkforimagesuperresolution
AT detianhuang tcsrlightweighttransformerandcnninteractionnetworkforimagesuperresolution