基于分类与最小卷积区域暗通道先验的水下图像恢复  

Underwater Image Restoration Based on Classification and Dark Channel Prior with Minimum Convolutional Area

作　　者：刘国栋 冯立辉[1] 卢继华[2] 崔建民[1] Liu Guodong;Feng Lihui;Lu Jihua;Cui Jianmin(School of Optics and Photonics,Beijing Institute of Technology,Beijing 100081,China;School of Integrated Circuits and Electronics,Beijing Institute of Technology,Beijing 100081,China)

机构地区：[1]北京理工大学光电学院,北京100081 [2]北京理工大学集成电路与电子学院,北京100081

出　　处：《激光与光电子学进展》2023年第4期56-66,共11页Laser & Optoelectronics Progress

基　　金：国家自然科学基金(62075012)。

摘　　要：为了解决水下图像在复杂水体中表现的画面模糊和颜色失真的问题,提出了一种基于HSV分类、CIELAB均衡与最小卷积区域暗通道先验(DCP)的水下图像恢复算法。基于H与S阈值将水下图像分为高饱和度失真图像、低饱和度失真图像及浅水图像等3类。分类后的水下图像分别经CIELAB均衡及自适应图像增强恢复,其中水下成像系统参数通过最小卷积区域DCP估计。实验结果表明,所提算法在图像恢复效果、评价质量和实时性指标上均优于对比算法,其中峰值信噪比和结构相似指数值分别平均提升了26.88%和17.3%,水下彩色图像质量评价值提升了4.3%。To address the issue of picture blur and color distortion in underwater images of complex water bodies,an underwater image restoration algorithm based on HSV classification,CIELAB equalization,and minimum convolution region dark channel prior(DCP)is proposed.By the thresholds of H and S,the underwater photos are separated into high saturation distortion,low saturation distortion,and shallow water images.Then,the underwater image is recovered using CIELAB equilibrium and adaptive image enhancement,where the system parameters of the categorized underwater image are estimated by minimum convolutional area DCP.The experimental findings demonstrate that the suggested solution is superior to the comparison algorithms in image restoration effect,evaluation quality,and realtime performance indicators.The average peak signaltonoise ratio and structural similarity values are increased by 26.88%and 17.3%on average,respectively,and the underwater image quality measurement value is increased by 4.3%.

关 键 词：海洋光学 图像阈值分类 颜色均衡 光学模型参数估计 峰值信噪比 水下彩色图像质量评价 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]