[1]宋 浩*,张 鸿.基于完全残差的双分支去雨网络[J].计算机技术与发展,2023,33(11):57-63.[doi:10. 3969 / j. issn. 1673-629X. 2023. 11. 009]
　SONG Hao*,ZHANG Hong.A Dual-branch De-raining Network Based on Fully Residual[J].,2023,33(11):57-63.[doi:10. 3969 / j. issn. 1673-629X. 2023. 11. 009]

点击复制

基于完全残差的双分支去雨网络()

《计算机技术与发展》[ISSN:1006-6977/CN:61-1281/TN]

卷:

33

期数:

2023年11期

页码:

57-63

栏目:

媒体计算

出版日期:

2023-11-10

文章信息/Info

Title:

A Dual-branch De-raining Network Based on Fully Residual

文章编号:

1673-629X(2023)11-0057-07

作者:

宋 浩¹; 2* ; 张 鸿¹; 2

1. 武汉科技大学 计算机科学与技术学院,湖北 武汉 430065;

2. 智能信息处理与实时工业系统湖北省重点实验室(武汉科技大学),湖北 武汉 430065

Author(s):

SONG Hao1; 2* ; ZHANG Hong1; 2

1. School of Computer Science and Technology,Wuhan University of Science and Technology,Wuhan 430065,China;

2. Hubei Province Key Laboratory of Intelligent Information Processing and Real-time Industrial System ( Wuhan University of Science and Technology) ,Wuhan 430065,China

关键词:

卷积神经网络; 单幅图像去雨; 多尺度学习; 完全残差; 可变形卷积

Keywords:

convolutional neural network; single image de-raining; multi-scale learning; fully residual; deformable convolution

分类号:

TP391. 41

DOI:

10. 3969 / j. issn. 1673-629X. 2023. 11. 009

摘要:

针对使用深度学习的单幅图像去雨算法会导致细节信息丢失的问题,提出了一个双分支去雨网络,包括雨痕提取分支和细节恢复分支,通过补全细节使去雨图像更接近真实图像。 雨痕提取分支的目的是完全提取出雨纹,通过构造特征金字塔来多尺度地学习雨纹信息,并在其中引入执行了全部身份映射的完全残差块来加强特征的重用和传播。 为充分利用上下文信息,采用可变形卷积在动态扩大感受野的同时避免了网格伪影的产生,最后输入雨图去除雨痕便得到了初步去雨图。 细节恢复分支需要产生细节特征图反馈给初步去雨图像来找回丢失的细节,使用轻量级的完全残差块捕捉特征信息,并用跳跃连接来连接完全残差块提供长距离的信息补偿。 实验结果表明,该网络在合成数据集 Rain100H 中比较RESCAN、SPANet 和 JDNet 等主流去雨方法,在 PSNR 和 SSIM 指标上分别至少提高了 0. 09 dB 和 0. 02,在真实数据集和自制数据集中的去雨效果和细节保留程度均优于对比方法。

Abstract:

Aiming at the problem that the single image de - raining algorithm using deep learning will lead to the loss of detailedinformation,a dual-branch de - raining network was proposed,including a rain streaks extraction branch and a detail recovery branch,which makes the de-raining image closer to the real image by completing the details. The purpose of the rain streaks extraction branch isto completely extract the rain streaks. A feature pyramid was constructed to learn the rain streaks information at multi - scale, andintroduce a fully residual block that performs all identity mapping in it to enhance feature re-usage and propagation. In order to take fulladvantage of context information, deformable convolution was used to dynamically expand the receptive field while avoiding thegeneration of grid artifacts,at the end input rain image take out rain streaks to?

obtain a preliminary de-raining image. The detail recoverybranch needs to generate a detail feature map to feed back to the preliminary de-raining image to retrieve the lost details,so a lightweightfully residual block was used to capture feature information, and a skip connections was used to connect the fully residual blocks toprovide long - distance information compensation. The experimental results show that the network improves the PSNR and SSIMindicators at least 0. 09 dB and 0. 02 respectively compared with the populous de-raining methods such as RESCAN,SPANet and JDNetin the synthetic dataset Rain100H,and the de-raining effect and the degree of detail retention are better than the comparison methods inboth the real dataset and the self-made dataset

常用功能