Scholarlink

<p>Underwater images often suffer from degradation such as color distortion, low contrast, and blurred details due to light absorption and scattering. Most existing underwater image enhancement (UIE) algorithms rely on a single network structure, limiting their ability to handle diverse degradation features effectively. To address this, we propose HiLoF-GAN, a novel Generative Adversarial Network (GAN) with high- and low-frequency feature separation. Our key innovation lies in the High and Low-Frequency Separation (HLFS) module within the generator, which explicitly extracts and processes high- and low-frequency features for finer detail enhancement. Additionally, we introduce a multi-color space loss function that integrates information from multiple color spaces to improve contrast and saturation more effectively. Extensive experiments on four benchmark datasets demonstrate that HiLoF-GAN outperforms state-of-the-art UIE methods in both qualitative and quantitative metrics, including PSNR, SSIM, LPIPS, and UCIQE, achieving significant improvements in detail preservation, color restoration, and overall visual quality.</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

HiLoF-GAN: high and low frequency information separation GAN for underwater image enhancement

  • 期刊论文

  • 发表日期：

    2025年9月26日

查看全文
  • Jiale Wang,
  • Yitao Liang,
  • Xuxu Yang,
  • Mengjuan Zhao,
  • Juan Xia

摘要

Underwater images often suffer from degradation such as color distortion, low contrast, and blurred details due to light absorption and scattering. Most existing underwater image enhancement (UIE) algorithms rely on a single network structure, limiting their ability to handle diverse degradation features effectively. To address this, we propose HiLoF-GAN, a novel Generative Adversarial Network (GAN) with high- and low-frequency feature separation. Our key innovation lies in the High and Low-Frequency Separation (HLFS) module within the generator, which explicitly extracts and processes high- and low-frequency features for finer detail enhancement. Additionally, we introduce a multi-color space loss function that integrates information from multiple color spaces to improve contrast and saturation more effectively. Extensive experiments on four benchmark datasets demonstrate that HiLoF-GAN outperforms state-of-the-art UIE methods in both qualitative and quantitative metrics, including PSNR, SSIM, LPIPS, and UCIQE, achieving significant improvements in detail preservation, color restoration, and overall visual quality.