Scholarlink

<p>Underwater imaging often suffers from color distortion and low contrast due to complex light absorption and scattering. This paper presents a hardware-agnostic enhancement framework that integrates adaptive optimization and advanced fusion techniques to address these issues. The process begins with adaptive color correction, followed by HSV transformation for targeted luminance processing. Principal Component Analysis (PCA) integrates the processed components into a globally enhanced image. Meanwhile, image decomposition refines details, and an improved Retinex algorithm enhances edge features. The Non-Subsampled Shearlet Transform (NSST) fuses the globally enhanced, detail-enhanced, and edge-enhanced images into a unified output, which is further optimized to produce the final high-quality result. Extensive experiments demonstrate superior performance across multiple metrics, including PCQI, UCIQE, UIQM, and IE, outperforming state-of-the-art methods. SIFT-based feature matching confirms enhanced feature preservation, underscoring the method’s potential in underwater exploration, photography, and robotic vision.</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

A robust framework for underwater image enhancement incorporating adaptive optimization and shearlet fusion

  • 期刊论文

  • 发表日期：

    2026年3月10日

查看全文
  • Yang Bai,
  • Zhe Li,
  • Libo Cheng

摘要

Underwater imaging often suffers from color distortion and low contrast due to complex light absorption and scattering. This paper presents a hardware-agnostic enhancement framework that integrates adaptive optimization and advanced fusion techniques to address these issues. The process begins with adaptive color correction, followed by HSV transformation for targeted luminance processing. Principal Component Analysis (PCA) integrates the processed components into a globally enhanced image. Meanwhile, image decomposition refines details, and an improved Retinex algorithm enhances edge features. The Non-Subsampled Shearlet Transform (NSST) fuses the globally enhanced, detail-enhanced, and edge-enhanced images into a unified output, which is further optimized to produce the final high-quality result. Extensive experiments demonstrate superior performance across multiple metrics, including PCQI, UCIQE, UIQM, and IE, outperforming state-of-the-art methods. SIFT-based feature matching confirms enhanced feature preservation, underscoring the method’s potential in underwater exploration, photography, and robotic vision.