EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Improved Face Image Super-Resolution Model Based on Generative Adversarial Network

Qingyu Liu iD, Yeguo Sun iD, Lei Chen, Lei Liu iD

Journal of Imaging · 19.05.2025 · 11 (5) · 163

DOI: 10.3390/jimaging11050163 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Journal of Imaging, online erschienen am 19.05.2025. Verfasst von Qingyu Liu und 3 weiteren Autor:innen. Beteiligte Einrichtung: School of Computer Science, Huainan Normal University, Huainan 232038, China. Die Arbeit wurde bisher 2-mal zitiert (Crossref) und nennt 32 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Key Projects of Natural Science Research in Anhui Colleges and Universities, University Natural Science Foundation of Anhui Province, Opening Foundation of State Key Laboratory of Cognitive Intelligence.

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Image super-resolution (SR) models based on the generative adversarial network (GAN) face challenges such as unnatural facial detail restoration and local blurring. This paper proposes an improved GAN-based model to address these issues. First, a Multi-scale Hybrid Attention Residual Block (MHARB) is designed, which dynamically enhances feature representation in critical face regions through dual-branch convolution and channel-spatial attention. Second, an Edge-guided Enhancement Block (EEB) is introduced, generating adaptive detail residuals by combining edge masks and channel attention to accurately recover high-frequency textures. Furthermore, a multi-scale discriminator with a weighted sub-discriminator loss is developed to balance global structural and local detail generation quality. Additionally, a phase-wise training strategy with dynamic adjustment of learning rate (Lr) and loss function weights is implemented to improve the realism of super-resolved face images. Experiments on the CelebA-HQ dataset demonstrate that the proposed model achieves a PSNR of 23.35 dB, a SSIM of 0.7424, and a LPIPS of 24.86, outperforming classical models and delivering superior visual quality in high-frequency regions. Notably, this model also surpasses the SwinIR model (PSNR: 23.28 dB → 23.35 dB, SSIM: 0.7340 → 0.7424, and LPIPS: 30.48 → 24.86), validating the effectiveness of the improved model and the training strategy in preserving facial details.

Schlagwörter

Stichwörter der Autor:innen

attention mechanismgenerative adversarial networkresidual modulesuper-resolution

Autor:innen und Institutionen

  • School of Computer Science, Huainan Normal University, Huainan 232038, China.
  • School of Finance and Mathematics, Huainan Normal University, Huainan 232038, China.

ORCID-Profile: Qingyu Liu, Yeguo Sun, Lei Liu

Förderung

  • Key Projects of Natural Science Research in Anhui Colleges and Universities
  • University Natural Science Foundation of Anhui Province
  • Opening Foundation of State Key Laboratory of Cognitive Intelligence
  • Program of Anhui Education Department

Zitieren

APA

Liu, Q., Sun, Y., Chen, L., & Liu, L. (2025). Improved Face Image Super-Resolution Model Based on Generative Adversarial Network. Journal of Imaging, 11(5), 163. https://doi.org/10.3390/jimaging11050163

Vancouver

Liu Q, Sun Y, Chen L, Liu L. Improved Face Image Super-Resolution Model Based on Generative Adversarial Network. Journal of Imaging. 2025;11(5):163. doi:10.3390/jimaging11050163

BibTeX
@article{Liu2025,
  author = {Liu, Qingyu and Sun, Yeguo and Chen, Lei and Liu, Lei},
  title = {Improved Face Image Super-Resolution Model Based on Generative Adversarial Network},
  journal = {Journal of Imaging},
  year = {2025},
  volume = {11},
  number = {5},
  pages = {163},
  doi = {10.3390/jimaging11050163}
}
Kostenlos starten →