Scholarlink

<p>Underwater imaging is crucial in various fields, yet it is significantly hindered by color distortion, light scattering and low visibility challenges. This paper introduces a no-reference image quality assessment (NR-IQA) method called (QUIQ) specifically designed for underwater images, focusing on a diverse array of features that capture the unique characteristics of underwater environments. We extracted 32 features, including sharpness measures in spatial and frequency domains, natural statistics from multiple scales, color statistics from the LAB color space, as well as entropy and local contrast metrics. These features were utilized to train a Gaussian Process Regression (GPR) model for predicting the objective quality of underwater images. Our results demonstrate strong correlations with subjective assessments and outperform existing underwater IQA methods, indicating the effectiveness of our approach in accurately capturing the complexities of underwater imaging. The code is available at <a href="https://github.com/mkarimid/QUIQ-Underwater-Image-Quality-Assessment">https://github.com/mkarimid/QUIQ-Underwater-Image-Quality-Assessment</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

QUIQ: quality-sensitive features for no-reference underwater image quality assessment

  • 期刊论文

  • 发表日期：

    2025年8月21日

查看全文
  • Maryam Karimi,
  • Meysam Ghalyani,
  • Seyede Fatemeh Noorani

摘要

Underwater imaging is crucial in various fields, yet it is significantly hindered by color distortion, light scattering and low visibility challenges. This paper introduces a no-reference image quality assessment (NR-IQA) method called (QUIQ) specifically designed for underwater images, focusing on a diverse array of features that capture the unique characteristics of underwater environments. We extracted 32 features, including sharpness measures in spatial and frequency domains, natural statistics from multiple scales, color statistics from the LAB color space, as well as entropy and local contrast metrics. These features were utilized to train a Gaussian Process Regression (GPR) model for predicting the objective quality of underwater images. Our results demonstrate strong correlations with subjective assessments and outperform existing underwater IQA methods, indicating the effectiveness of our approach in accurately capturing the complexities of underwater imaging. The code is available at https://github.com/mkarimid/QUIQ-Underwater-Image-Quality-Assessment.