EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

EHNet: Efficient Hybrid Network with Dual Attention for Image Deblurring

Quoc-Thien Ho iD, Minh-Thien Duong iD, Seongsoo Lee iD, Min-Cheol Hong iD

Sensors · 10.10.2024 · 24 (20) · 6545

DOI: 10.3390/s24206545 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Sensors, online erschienen am 10.10.2024. Verfasst von Quoc-Thien Ho und 3 weiteren Autor:innen. Beteiligte Einrichtung: Department of Information and Telecommunication Engineering, Soongsil University, Seoul 06978, Republic of Korea. Die Arbeit wurde bisher 10-mal zitiert (Crossref) und nennt 57 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: MOTIE, KEIT, KIAT, MOTIE.

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

The motion of an object or camera platform makes the acquired image blurred. This degradation is a major reason to obtain a poor-quality image from an imaging sensor. Therefore, developing an efficient deep-learning-based image processing method to remove the blur artifact is desirable. Deep learning has recently demonstrated significant efficacy in image deblurring, primarily through convolutional neural networks (CNNs) and Transformers. However, the limited receptive fields of CNNs restrict their ability to capture long-range structural dependencies. In contrast, Transformers excel at modeling these dependencies, but they are computationally expensive for high-resolution inputs and lack the appropriate inductive bias. To overcome these challenges, we propose an Efficient Hybrid Network (EHNet) that employs CNN encoders for local feature extraction and Transformer decoders with a dual-attention module to capture spatial and channel-wise dependencies. This synergy facilitates the acquisition of rich contextual information for high-quality image deblurring. Additionally, we introduce the Simple Feature-Embedding Module (SFEM) to replace the pointwise and depthwise convolutions to generate simplified embedding features in the self-attention mechanism. This innovation substantially reduces computational complexity and memory usage while maintaining overall performance. Finally, through comprehensive experiments, our compact model yields promising quantitative and qualitative results for image deblurring on various benchmark datasets.

Schlagwörter

Stichwörter der Autor:innen

Transformerconvolution neural networksdual attention modulehybrid architectureimage deblurringmotion blur

Autor:innen und Institutionen

  • Department of Information and Telecommunication Engineering, Soongsil University, Seoul 06978, Republic of Korea.
  • Department of Automatic Control, Ho Chi Minh City University of Technology and Education, Ho Chi Minh City 70000, Vietnam.
  • Department of Intelligent Semiconductor, Soongsil University, Seoul 06978, Republic of Korea.
  • School of Electronic Engineering, Soongsil University, Seoul 06978, Republic of Korea.

ORCID-Profile: Quoc-Thien Ho, Minh-Thien Duong, Seongsoo Lee, Min-Cheol Hong

Förderung

  • MOTIE, KEIT
  • KIAT
  • MOTIE
  • Ministry of Trade, Industry, and Energy (MOTIE)
  • Korea Evaluation Institute of Industrial Technology (KEIT)
  • Korea Institute for Advancement of Technology (KIAT)
  • Innovative Human Resource Development for Local Intellectualization Program
  • IC Design Education Center (IDEC)

Zitieren

APA

Ho, Q. T., Duong, M. T., Lee, S., & Hong, M. C. (2024). EHNet: Efficient Hybrid Network with Dual Attention for Image Deblurring. Sensors, 24(20), 6545. https://doi.org/10.3390/s24206545

Vancouver

Ho QT, Duong MT, Lee S, Hong MC. EHNet: Efficient Hybrid Network with Dual Attention for Image Deblurring. Sensors. 2024;24(20):6545. doi:10.3390/s24206545

BibTeX
@article{Ho2024,
  author = {Ho, Quoc-Thien and Duong, Minh-Thien and Lee, Seongsoo and Hong, Min-Cheol},
  title = {EHNet: Efficient Hybrid Network with Dual Attention for Image Deblurring},
  journal = {Sensors},
  year = {2024},
  volume = {24},
  number = {20},
  pages = {6545},
  doi = {10.3390/s24206545}
}
Kostenlos starten →