Scholarlink

<p>Underwater images often suffer from degradation due to light absorption and scattering, leading to issues such as blurriness, low contrast, and color distortion. To address these challenges, we advance the dark channel low-rank prior theory for the first time and propose a novel single underwater image restoration method based on it. Our method, called DCLR, comprises four key steps: color correction preprocessing, transmission estimation through matrix completion, water light estimation using image segmentation, and scene radiance recovery. By leveraging the low-rank nature of dark channels, DCLR enhances transmission estimation accuracy without relying on soft matting or guided filters, thereby improving descattering capabilities. Experiments on multiple datasets demonstrate that DCLR effectively improves color deviations and significantly enhances underwater image restoration outcomes. The source code and datasets will be available at: <a href="https://github.com/WYUlin16/WYUlin%20%28DOI:10.5281/zenodo.13825112)">https://github.com/WYUlin16/WYUlin (DOI:10.5281/zenodo.13825112)</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

Dark Channel Low-Rank Prior for Enhanced Single Underwater Image Restoration

  • 期刊论文

  • 发表日期：

    2025年6月30日

查看全文
  • Yulin Wang,
  • Zheng Liang,
  • Zetian Mi,
  • Jiqing Zhang,
  • Xianping Fu,
  • Yujia Wang

摘要

Underwater images often suffer from degradation due to light absorption and scattering, leading to issues such as blurriness, low contrast, and color distortion. To address these challenges, we advance the dark channel low-rank prior theory for the first time and propose a novel single underwater image restoration method based on it. Our method, called DCLR, comprises four key steps: color correction preprocessing, transmission estimation through matrix completion, water light estimation using image segmentation, and scene radiance recovery. By leveraging the low-rank nature of dark channels, DCLR enhances transmission estimation accuracy without relying on soft matting or guided filters, thereby improving descattering capabilities. Experiments on multiple datasets demonstrate that DCLR effectively improves color deviations and significantly enhances underwater image restoration outcomes. The source code and datasets will be available at: https://github.com/WYUlin16/WYUlin (DOI:10.5281/zenodo.13825112).