现代电子技术2024，Vol.47Issue(19)：55-61,7.DOI:10.16652/j.issn.1004-373x.2024.19.009
基于零参考深度曲线估计的水下图像增强算法
Underwater image enhancement algorithm based on zero reference depth curve estimation
摘要
Abstract
Underwater photography is challenged by optical distortions caused by absorption and scattering of water.These distortions manifest as color aberrations,image blurring and reduced contrast in underwater scenes.In view of the above,a no-reference underwater image enhancement algorithm is proposed.On the basis of the convolutional neural network(CNN),the algorithm realizes underwater image enhancement by combining with curve estimation.The shallow characteristics of the image are retained by the convolution layer first,and then the detailed information of image features is compensated by connecting dense residual blocks.Finally,the extracted feature information is subjected to curve estimation,and the pixel value is adjusted dynamically,so as to obtain a clear image.In the process of network training,a set of no-reference loss functions are used to drive the network learning,which can improve the image quality without the need of paired data.The model performance is evaluated and tested on the public data set.Comparison analysis against other prominent enhancement methods demonstrates the superiority of the proposed algorithm.It's PSNR(peak signal-to-noise ratio)and SSIM(structural similarity index measure)reach 23.544 and 0.830,respectively,surpassing the second-best algorithm by 10.02％ and 3.88％,respectively.关键词
水下图像增强/曲线估计/无参考损失函数/残差网络/深度学习/跳跃连接Key words
underwater image enhancement/curve estimation/no-reference loss function/residual network/deep learning/skip connection分类
电子信息工程引用本文复制引用
冯岩,张文鹏,刘劲芸,安永丽..基于零参考深度曲线估计的水下图像增强算法[J].现代电子技术,2024,47(19):55-61,7.基金项目
国家科技部重点研发专项(2017YFE0135700) （2017YFE0135700）
河北省高层次人才工程项目(A201903011) （A201903011）
河北省自然科学基金项目(F2018209358) （F2018209358）
