基于非线性信道先验的水下盲图像复原方法  

Underwater blind image restoration method based on the nonlinear channel prior

作　　者：黄浩乾 马惊天 武鹏 郑康健 晋云飞 邱榕 HUANG Haoqian;MA Jingtian;WU Peng;ZHENG Kangjian;JIN Yunfei;QIU Rong(Hohai University College of Energy and Electrical Engineering,Nanjing 211100,China;Nanjing Huiying Electronic Technology Co.,Ltd,Nanjing 210000,China)

机构地区：[1]河海大学能源与电气学院,南京211100 [2]南京荟英电子科技有限公司,南京210000

出　　处：《中国惯性技术学报》2023年第7期665-673,共9页Journal of Chinese Inertial Technology

基　　金：国家自然科学基金(61703098);江苏省自然科学基金面上项目(BK20221500)。

摘　　要：为了将水流干扰下的运动模糊图像恢复成清晰图像,提出基于非线性信道先验的水下盲图像复原方法。第一,利用差异通道增益对红色通道进行初始补偿,估计出精确的环境背景光;第二,求取红通道的粗略传输图,对初始传输图进行导向滤波处理得到精准的透射率;第三,在红、绿、蓝三通道中附加各自通道的环境背景光来统一上述透射率,利用补偿屏蔽的方式合并得到去雾的水下盲图像;第四,基于非线性信道先验获得去雾水下盲图像的初始复原图像,将初始复原图像、模糊核以及潜在清晰图像经过多次迭代计算得到最终的水下复原图像。实验结果表明,与水下暗通道先验算法、水下retinex增强以及水下归一化总变分方法相比,所提方法在Ancuti数据集上的水下图像评价指标值分别提升了21.7%、10.9%、26.5%,对水流干扰下的运动模糊图像有较好的复原效果。In order to restore the underwater motion blurred image to a clear image,a blind underwater image restoration method based on nonlinear channel prior is proposed.Firstly,the red channel is initially compensated by the differential channel gain to estimate the accurate background light.Secondly,the rough transmission map of the red channel is obtained,and the initial transmission map is filtered to acquire the accurate transmittance.Thirdly,the environmental background light of each channel in the three channels are contributed to unify the transmissivity,and utilize the compensation shielding method to get hold of the defogging underwater blind image.Fourthly,based on the nonlinear channel prior,the initial restored image of the defogged underwater blind image is obtained.The initial restored image,fuzzy kernel,and potentially clear image are iteratively calculated multiple times to obtain the final underwater restored image.The experimental results show that compared with the underwater dark channel prior(UDCP),underwater retinex enhancement(URTE)and underwater normalized total variation(UNTV)methods,the underwater image evaluation index values of the proposed method on the Ancuti dataset have been improved by 21.7%,10.9%,and 26.5%,respectively,which has a good restoration effect on motion blurred images under water flow interference.

关 键 词：水下盲图像复原 非线性信道先验 潜在清晰图像 模糊核 

分 类 号：TJ630[兵器科学与技术—武器系统与运用工程] TP391[自动化与计算机技术—计算机应用技术]