Scholarlink

<p>Enhancing underwater images and videos is essential for a number of applications, including underwater autonomous vehicles, marine ecosystems, and submarine archaeology. Due to water turbidity, light scattering, and absorption, these conditions present additional challenges such as color distortion, low brightness, and blurring. While equalization of histograms and Retinex-based techniques have been employed to improve underwater photos, they are insufficient to address these problems holistically. Over the past few years, deep learning (DL) techniquesin particular, Convolutional Neural Networks, and Generative Adversarial Networks have been increasingly popular for improving underwater photos and videos. By restoring missing data, boosting contrast, color weighing, and minimizing noise, these models may enhance image quality and discover intricate patterns from massive datasets. Hybrid techniques that integrate deep learning with physical models, i.e., scattering and absorption models, hold promise in addressing actual underwater challenges. Nevertheless, not withstanding these developments, there are some challenges that are yet to be addressed, which include high computational expense, requirements for large amounts of labelled datasets, and complexity in real-time processing. The future should entail the development of real-time enhancement models, multi-modal enhancement strategies, and hybrid techniques, and also addressing the limitations of datasets. On-going development in these fields will enhance underwater imaging systems for more efficient use in marine biology, archaeology, environmental monitoring, and autonomous navigation.</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

Advances in underwater image and video enhancement systematic review of conventional and deep learning-based method

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年5月8日

查看全文
  • M. Thanjaivadivel,
  • Rajasekaran Thangaraj

摘要

Enhancing underwater images and videos is essential for a number of applications, including underwater autonomous vehicles, marine ecosystems, and submarine archaeology. Due to water turbidity, light scattering, and absorption, these conditions present additional challenges such as color distortion, low brightness, and blurring. While equalization of histograms and Retinex-based techniques have been employed to improve underwater photos, they are insufficient to address these problems holistically. Over the past few years, deep learning (DL) techniquesin particular, Convolutional Neural Networks, and Generative Adversarial Networks have been increasingly popular for improving underwater photos and videos. By restoring missing data, boosting contrast, color weighing, and minimizing noise, these models may enhance image quality and discover intricate patterns from massive datasets. Hybrid techniques that integrate deep learning with physical models, i.e., scattering and absorption models, hold promise in addressing actual underwater challenges. Nevertheless, not withstanding these developments, there are some challenges that are yet to be addressed, which include high computational expense, requirements for large amounts of labelled datasets, and complexity in real-time processing. The future should entail the development of real-time enhancement models, multi-modal enhancement strategies, and hybrid techniques, and also addressing the limitations of datasets. On-going development in these fields will enhance underwater imaging systems for more efficient use in marine biology, archaeology, environmental monitoring, and autonomous navigation.