Article Navigation > homepage  Archive  2026  No.1
> DOI:10.16366/j.cnki.1000-2367.2024.12.11.0004

R2PixGAN:an efficient new method for denoising oracle bone topographies

Number of views:10
  • 分享到:

摘要:

随着技术的发展,深度学习在图像去噪领域取得了显著的效果.然而,由于甲骨拓片往往同时包括各种噪声,现有的去噪模型无法适应甲骨文独特的字形和复杂的文字背景.针对上述挑战,提出了一种基于条件对抗网络(Pix2Pix)的图像去噪方法 R2PixGAN.在该方法中,生成器采用了 R2U-Net模型,该模型不仅保留了传统U-Net在特征提取方面的优势,还通过引人循环神经网络(RNN)结构,进一步提升了图像重建能力,同时增强了去噪效果.此外,还将感知损失纳人模型,以更好地保留原始图像的关键细节和特征.实验结果表明,R2PixGAN在PSNR和SSIMI分数方面优于对比实验,图像去噪效果得到了明显的提升.

Deep learning has shown strong effects in image denoising. However, existing models often fail to handle ora-cle bone rubbings which contain complex noise and unique character structures. To solve this, we propose R2PixGAN, a denoising method based on Pix2Pix. It uses an R2U-Net as the generator which keeps the benefits of U-Net and adds RNN to im-prove image rebuilding and denoising, We also include a perceptual loss to keep key details. Tests show that R2PixGAN gets higher PSNR and SSIM scores than other methods, proving its better performance.

作者: 

王士斌,王宇,喻琪,刘栋,闫娟

Wang Shibin,Wang Yu, Yu Qi,Liu Dong,Yan Juan

机构地区:

河南师范大学 a.计算机与信息工程学院;b.河南省教育人工智能与个性化学习重点实验室;c.甲骨智能计算实验室

引用本文:

王士斌,王宇,喻琪等。R2PixGAN:一种高效的甲骨文拓片去噪新方法[J].河南师范大学学报(自然科学版),2026,54(1):1-7.(Wang Shibin, Wang Yu, Yu Qi,et al.R2PixGAN:an efficient new method for denoising oracle bone topographies[J].Journal of Henan Normal University(Natural Science Edition),2026,54(1): 1-7.DOI:10.16366/j.cnki.1000-2367.2024.12.11.0004.)

基金:

国家自然科学基金;河南省高等学校重点科研项目

关键词:

图像去噪;甲骨文;感知损失

image denoising; oracle bone inscriptions; perceptual loss

分类号:

TP399


R2PixGAN:一种高效的甲骨文拓片去噪新方法.pdf


Links
更多+
  • Henan Normal University
  • CNKI
  • Elsevier
  • Springer
  • Scopus Preview
  • Web of Science