融合红外与可见光图像的三分支网络
Three Branch Network for Fusion of Infrared and Visible Images
周丁 1王逊 1黄树成1
作者信息
- 1. 江苏科技大学计算机学院 镇江 212100
- 折叠
摘要
针对红外图像与可见光图像的融合,现有的图像融合任务基本上是基于普通的自编码器网络,忽略了对图像重要特征的提取.针对这一问题,提出了一种独特的自编码器网络,网络特点在于拥有三个分支,分别是纹理分支、内容分支、基础分支.在编码器中,通过三个分支提取红外图像与可见光图像的重要特征,然后融合两组特征得到融合特征.解码器对融合特征解码得到融合的图像.将此方法应用于图像融合实验,取得了较好的实验结果.
Abstract
For the fusion of infrared image and visible image,the existing image fusion task is basically based on the common autoencoder network,ignoring the extraction of important features of the image.To solve this problem,a unique autoencoder net-work is proposed,which has three branches,which are texture branch,content branch and base branch.In the encoder,the impor-tant features of infrared image and visible image are extracted by three branches,and then the fusion features are obtained by fusing the two groups of features.The decoder decodes the fusion feature to obtain the fusion image.This method is applied to the image fu-sion experiment and good experimental results are obtained.
关键词
图像融合/红外图像与可见光图像/特征选择/三分支自编码器Key words
image fusion/infrared image and visible image/feature selection/three branch autoencoder引用本文复制引用
出版年
2024
国家科技期刊平台
NSTL
万方数据