Scholarlink

<p>Sandstorm images are often plagued by color distortions, reduced contrast, and blurred details, posing challenges for advanced vision tasks. To address these issues, we propose a novel color-guided spatial-frequency fusion network (CSF) for sandstorm image enhancement. The CSF framework comprises a color-guided module (CGM) for correcting color distortions and a dual-domain feature fusion module (DFFM) that integrates spatial and frequency domain features. CGM dynamically adjusts color information across RGB channels, facilitating preliminary color correction. DFFM extracts multi-scale spatial features and global frequency components, enabling the network to learn robust and discriminative representations. Experimental results on both synthetic and real-world sandstorm datasets demonstrate that CSF outperforms state-of-the-art methods in terms of qualitative and quantitative performance, highlighting its effectiveness for enhancing sandstorm images. Our code and datasets can be available at <a href="https://github.com/cvandpr/CSF">https://github.com/cvandpr/CSF</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

Enhancing sandstorm images via color-guided spatial-frequency fusion network

  • 期刊论文

  • 发表日期：

    2025年4月15日

查看全文
  • Zhengwei Guo,
  • Bo Wang

摘要

Sandstorm images are often plagued by color distortions, reduced contrast, and blurred details, posing challenges for advanced vision tasks. To address these issues, we propose a novel color-guided spatial-frequency fusion network (CSF) for sandstorm image enhancement. The CSF framework comprises a color-guided module (CGM) for correcting color distortions and a dual-domain feature fusion module (DFFM) that integrates spatial and frequency domain features. CGM dynamically adjusts color information across RGB channels, facilitating preliminary color correction. DFFM extracts multi-scale spatial features and global frequency components, enabling the network to learn robust and discriminative representations. Experimental results on both synthetic and real-world sandstorm datasets demonstrate that CSF outperforms state-of-the-art methods in terms of qualitative and quantitative performance, highlighting its effectiveness for enhancing sandstorm images. Our code and datasets can be available at https://github.com/cvandpr/CSF.