当前位置: X-MOL 学术Optik › 论文详情
Our official English website, www.x-mol.net, welcomes your feedback! (Note: you will need to create a separate account there.)
The underwater polarization dehazing imaging with a lightweight convolutional neural network
Optik Pub Date : 2021-11-26 , DOI: 10.1016/j.ijleo.2021.168381
Qiming Ren 1 , Yanfa Xiang 1 , Guochen Wang 1 , Jie Gao 1 , Yan Wu 1 , Rui-Pin Chen 1
Affiliation  

The scattering and absorption of particles in underwater environment seriously affect the quality of underwater images, resulting in reduced contrast and imaging quality. In this work, the underwater active polarization dehazing imaging based on the deep learning model is studied. A modified lightweight dehazing convolutional neural network (CNN) model with four input channels is designed by combining both the advantages of the deep learning and polarization dehazing imaging technology. The lightweight CNN is trained and tested with the images of different polarization components (00, 450, 900 linear polarization and circular polarization) in different turbidity underwater environments. The experimental results show that this method can rapidly achieve the better dehazing imaging effect than that of conventional dehazing methods.



中文翻译:

基于轻量级卷积神经网络的水下偏振去雾成像

水下环境中粒子的散射和吸收严重影响水下图像的质量,导致对比度和成像质量下降。本文研究了基于深度学习模型的水下主动偏振去雾成像。结合深度学习和偏振去雾成像技术的优点,设计了一种改进的具有四个输入通道的轻量级去雾卷积神经网络(CNN)模型。轻量级 CNN 使用不同偏振分量 (0 0 , 45 0 , 90 0线偏振和圆偏振)在不同浊度的水下环境中。实验结果表明,与传统的去雾方法相比,该方法可以快速达到更好的去雾成像效果。

更新日期:2021-12-01
down
wechat
bug