摘要
传统红外与可见光融合图像在复杂环境下存在细节缺失,特征单一导致目标模糊等问题,本文提出一种基于卷积神经网络结合非下采样轮廓波变换(non-subsampled contourlet transform,NSCT)的红外与可见光图像进行融合的方法。首先,通过卷积神经网络提取红外与可见光目标特征信息,同时利用NSCT对源图像进行多尺度分解,得到源图像的高频系数与低频系数;其次,结合目标特征图利用自适应模糊逻辑与局部方差对比度分别对源图像高频子带与低频子带进行融合;最后,通过逆NSCT变换得到融合图像并与其他5种传统算法进行对比;实验结果表明,本文方法在多个客观评价指标上均有所提高。
Traditional infrared and visible fused images suffer from missing details and blurred targets owing to single features in complex environments.This study presents a method for fusing infrared and visible images based on a convolution neural network(CNN)combined with a non-subsampled contourlet transform(NSCT).Firstly,the infrared and visible target feature information is extracted by CNN,and the source image is decomposed by the NSCT at multiple scales to obtain its high-frequency coefficients and lowfrequency coefficients.Secondly,the high-frequency sub-bands and low-frequency sub-bands of the source image are fused separately using adaptive fuzzy logic and local variance contrast in combination with the target feature image.Finally,the fused image is obtained by inverse NSCT transformation.We conducted a comparative analysis with five other traditional algorithms.The experimental results show that the proposed method performs better in several objective evaluation indicators.
作者
曹宇彤
宦克为
薛超
韩丰地
李向阳
陈笑
CAO Yutong;HUAN Kewei;XUE Chao;HAN Fengdi;LI Xiangyang;CHEN Xiao(College of Physics,Changchun University of Science and Technology,Changchun 130022,China)
出处
《红外技术》
CSCD
北大核心
2023年第4期378-385,共8页
Infrared Technology
基金
国家自然科学基金(61905026)
吉林省科技发展计划项目(20210101158JC)。
