Scholarlink

<p>Dust storms significantly degrade image quality, posing challenges for computer vision applications. Existing methods face limitations in handling complex degradations in dusty images. This paper proposes a novel dusty image restoration network, PGAC-Net (Polarization-Gabor Attention Correction Network), which integrates polarization-inspired Gabor features, gradient domain enhancement, and scattering-aware guided correction. Employing an encoder-decoder architecture, the network integrates two core innovative modules: the Scattering-Aware Dynamic Attention Module (SADA) enables channel-differentiated degradation estimation and precise color correction, while the Direction-Gradient Feature Merging Module (PGFM) enhances detail and edge restoration capabilities. Comprehensive validation using synthetic datasets and real dust storm images demonstrates that PGAC-Net outperforms existing methods in both objective evaluation metrics (PSNR, SSIM, etc.) and subjective visual quality, significantly enhancing image clarity and contrast. Our code can be available at <a href="https://github.com/15771874423-sketch/PGAC-Net.git">https://github.com/15771874423-sketch/PGAC-Net.git</a>.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

PGAC-Net: sandstorm image restoration network based on scattering-aware guidance and gabor feature enhancement

  • 期刊论文

  • 发表日期：

    2026年4月17日

查看全文
  • Hongxia Niu,
  • Gao Wang,
  • Tao Hou

摘要

Dust storms significantly degrade image quality, posing challenges for computer vision applications. Existing methods face limitations in handling complex degradations in dusty images. This paper proposes a novel dusty image restoration network, PGAC-Net (Polarization-Gabor Attention Correction Network), which integrates polarization-inspired Gabor features, gradient domain enhancement, and scattering-aware guided correction. Employing an encoder-decoder architecture, the network integrates two core innovative modules: the Scattering-Aware Dynamic Attention Module (SADA) enables channel-differentiated degradation estimation and precise color correction, while the Direction-Gradient Feature Merging Module (PGFM) enhances detail and edge restoration capabilities. Comprehensive validation using synthetic datasets and real dust storm images demonstrates that PGAC-Net outperforms existing methods in both objective evaluation metrics (PSNR, SSIM, etc.) and subjective visual quality, significantly enhancing image clarity and contrast. Our code can be available at https://github.com/15771874423-sketch/PGAC-Net.git.