现代电子技术2024，Vol.47Issue(23)：15-21,7.DOI:10.16652/j.issn.1004-373x.2024.23.003
基于积极概率分布引导的水下图像增强网络
Underwater image enhancement network based on guidance of positive probability distribution
摘要
Abstract
The underwater environment is complex and changeable,which makes it difficult for the existing algorithms to restore the degraded underwater images stably,so an underwater image enhancement network based on the guidance of positive probability distribution is proposed,that is,by constructing a probability guidance framework of positive samples and estimating their characteristic probability distributions from mixed samples,the network is guided to restore the degraded underwater images.Firstly,a multi-branch information extraction architecture is proposed to obtain the multi-spatial features of input features,and enhance the image feature distributions in the aspects of space,pixels and channels.Secondly,the feature distribution is obtained by combining conditional variational autoencoder(CVAE)with adaptive instance normalization,and the existing style transfer algorithm is improved to enhance image color and contrast.Finally,the robustness of the network and the image quality are further improved by combining multiple loss functions.The experimental results show that the output image of the proposed method has clear and uniform color,and its qualitative and quantitative indicators on several data sets are superior to those of the classical and novel algorithms,so it is of great significance to real transactions of marine.关键词
水下图像增强/概率分布/条件变分自动编码器/自适应实例归一化/注意力机制/色偏校正Key words
underwater image enhancement/probability distribution/CVAE/adaptive case normalization/attention mechanism/color cast correction分类
电子信息工程引用本文复制引用
朱立忠,王雅鑫,郭宝仁..基于积极概率分布引导的水下图像增强网络[J].现代电子技术,2024,47(23):15-21,7.基金项目
国家重点研发计划(2017YFC0821001-2) （2017YFC0821001-2）
