基于AGF和CNN的红外与可见光图像融合
Infrared and visible image fusion based on AGF and CNN
杨艳春 1杨万轩 1雷慧云1
作者信息
- 1. 兰州交通大学电子与信息工程学院,甘肃兰州 730070
- 折叠
摘要
针对红外与可见光图像融合中出现的边缘模糊和细节丢失等问题,本文提出了 一种基于交替引导滤波器(AGF)与掩膜引导卷积神经网络(CNN)的融合算法.首先,将源图像通过交替引导滤波分解为基础层与细节层;然后,将基础层通过能量属性的融合规则得到基础融合图像,细节层在基于掩膜引导的损失函数的指导下,通过卷积神经网络得到融合后的细节图像;最后,将基础融合图像与细节融合图像相加得到最终融合图像;实验结果表明,本文方法能够在突出显著热目标的同时保留丰富的背景边缘纹理信息,在客观评价指标上相较对比方法取得了更好的效果,证明了本文算法的优越性.
Abstract
Aiming at the problems of edge blurring and detail loss in the fusion of infrared and visible image,this pa-per proposes a fusion algorithm based on alternating guided filter(AGF)and mask-guided convolutional neural net-work(CNN).Firstly,the source images is decomposed into a base layer and a detail layer by alternating guided filte-ring.Then,the base layer is passed through the fusion rule with energy attributes to get the base fusion image,and the detail layer is guided by the loss function based on the mask guidance to get the fused detail image by convolutional neural network.Finally,the base fusion image and the detail fusion image are summed to generate the final fused im-age.The experimental results demonstrate that the proposed method effectively retains abundant background edge tex-ture information while highlighting significant thermal targets,and achieves better results in objective evaluation met-rics compared with the comparison methods,which proves the superiority of the proposed algorithm.
关键词
图像处理/红外与可见光图像/交替引导滤波/卷积神经网络/图像融合Key words
image processing/infrared and visible image/alternating guided filter/convolutional neural network/image fusion引用本文复制引用
基金项目
长江学者和创新团队发展计划资助项目(IRT_16R36)
国家自然科学基金项目(62067006)
甘肃省科技计划项目(18JR3RA104)
甘肃省高等学校产业支撑计划项目(2020C-19)
兰州市科技计划项目(2019-4-49)
甘肃省教育厅青年博士基金项目(2022QB-067)
甘肃省自然科学基金项目(23JRRA847)
甘肃省自然科学基金项目(21JR7RA300)
兰州交通大学天佑创新团队项目(TY202003)
兰州交通大学—天津大学联合创新基金项目(2021052)
出版年
2024
NETL
NSTL
万方数据