Research on Hardware Acceleration of Dehazing Network Based on Deep Learning
International Journal of Computer Science and Information Technology · 23.01.2025 · 5 (1) · 69-87
DOI: 10.62051/ijcsit.v5n1.07 ↗Kurz gefasst
Fachartikel in der Zeitschrift International Journal of Computer Science and Information Technology, online erschienen am 23.01.2025. Verfasst von Xianlong Zheng. Die Arbeit wurde bisher 1-mal zitiert (Crossref) und nennt 15 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY-NC 4.0).
Abstract (Originalsprache)
Computer vision devices play a crucial role in both industrial and everyday life. However, when capturing images in hazy weather, the performance of these devices can be severely affected, resulting in a significant decrease in image quality. Therefore, effective image dehazing has become a necessary and urgent task. In recent years, with the rapid development of deep learning technology, significant progress has also been made in the field of image dehazing. However, despite the excellent performance of existing models in improving dehazing effects, they often overlook issues of computational efficiency and resource consumption. The high computational cost not only limits the widespread application of these advanced methods in practical scenarios, but also increases the difficulty and cost of hardware deployment. In response to the above challenges, we propose a novel lightweight image dehazing network called LDM-Net (Lightweight Dehazing Module Network). This network aims to reduce model complexity by optimizing structural design, while introducing attention mechanisms to enhance its generalization ability, thus better adapting to image dehazing needs in different environments. In order to verify the actual effectiveness and potential value of LDM-Net, we successfully deployed this network on various hardware platforms and conducted comprehensive testing and evaluation. The experimental results show that compared to several leading image dehazing solutions in the current market, LDM-Net can achieve better or at least equivalent dehazing performance while maintaining lower computational resource consumption. Specifically, it can approach or even surpass the performance of other advanced lightweight image dehazing methods in multiple key indicators, including but not limited to peak signal-to-noise ratio (PSNR), structural similarity index (SSIM), and visual perception evaluation. This achievement not only demonstrates the strong competitiveness of LDM-Net as an efficient and low-energy solution, but also points out a new direction for the research and development of image dehazing technology in the future. In addition, considering its good scalability and flexibility, LDM-Net is expected to be widely used in smart city construction, auto drive system and other fields, further promoting the progress and development of related industries.
Zitieren
APA
Zheng, X. (2025). Research on Hardware Acceleration of Dehazing Network Based on Deep Learning. International Journal of Computer Science and Information Technology, 5(1), 69-87. https://doi.org/10.62051/ijcsit.v5n1.07
Vancouver
Zheng X. Research on Hardware Acceleration of Dehazing Network Based on Deep Learning. International Journal of Computer Science and Information Technology. 2025;5(1):69-87. doi:10.62051/ijcsit.v5n1.07
BibTeX
@article{Zheng2025,
author = {Zheng, Xianlong},
title = {Research on Hardware Acceleration of Dehazing Network Based on Deep Learning},
journal = {International Journal of Computer Science and Information Technology},
year = {2025},
volume = {5},
number = {1},
pages = {69--87},
doi = {10.62051/ijcsit.v5n1.07}
}
EUVIMED