深度学习驱动的水下图像增强与复原研究进展

Translated title of the contribution: Research Progress of Deep Learning Driven Underwater Image Enhancement and Restoration

丛润民, 张禹墨, 张晨, 李重仪, 赵耀

Research output: Journal PublicationsJournal Article (refereed)peer-review

Abstract

水下图像是水下信息的重要载体和呈现方式,对海洋资源的探索、开发、利用具有至关重要的作用。然而,由于客观成像环境和设备的限制,水下图像质量总是差强人意,具有对比度低、细节模糊、颜色偏差等退化现象,严重制约相关领域的发展。因此,如何通过后期算法对退化的水下图像进行增强和复原越来越受到学者们的关注。近些年,随着深度学习技术的快速发展,基于深度学习的水下图像增强与复原技术取得了巨大进展。为了更加全面、立体地对现有方法进行梳理与归纳,紧跟最新研究进展,本文着重对深度学习驱动的水下图像增强与复原的方法和模型进行介绍,详细整理现有的水下图像数据集,分析现有基于深度学习方法的关键问题,并对未来发展方向进行展望。
Underwater images are important carrier and representation of underwater information, which is crucial for ocean resources' exploration, development, and utilization. However, due to the limitations of objective imaging environment and imaging equipment, the quality of underwater image is always unsatisfactory, such as low contrast, blurred details, and color deviations, seriously restricting the development of related research areas. Therefore, how to enhance and restore degraded underwater images through the designed post-processing algorithms has attracted more and more attention. In recent years, with the rapid development of deep learning strategy, deep learning-based underwater image enhancement and restoration technology has also made great progresses. In order to comprehensively summarize the existing methods and keep up with the latest research progress, this paper introduces the methods and models of deep learning driven underwater image enhancement and restoration technology, summarizes the existing underwater image datasets in detail, analyzes the key is sues of the existing deep learning-based methods, and prospects for the future development directions.
Translated title of the contributionResearch Progress of Deep Learning Driven Underwater Image Enhancement and Restoration
Original languageChinese (Simplified)
Pages (from-to)1377-1389
Number of pages13
Journal信号处理 = Journal of Signal Processing
Volume36
Issue number9
DOIs
Publication statusPublished - Sept 2020
Externally publishedYes

Keywords

  • 水下图像
  • 质量退化
  • 图像增强与复原
  • 深度学习
  • 水下图像增强与复原数据库
  • underwater image
  • quality degradation
  • image enhancement and restoration
  • deep learning
  • underwater image enhancement and restoration dataset

Fingerprint

Dive into the research topics of 'Research Progress of Deep Learning Driven Underwater Image Enhancement and Restoration'. Together they form a unique fingerprint.

Cite this