基于NSCT域平均梯度能量驱动红外与可见光图像融合算法  

Infrared and Visible Image Fusion based on AverageEnergy Driving in NSCT Domain

作　　者：贾鑫 JIA Xin(School of Mechanical Engineering,Beihua University,Jilin City 132021,China)

机构地区：[1]北华大学机械工程学院,吉林吉林132021

出　　处：《吉林化工学院学报》2023年第3期86-92,共7页Journal of Jilin Institute of Chemical Technology

摘　　要：针对红外与可见光图像融合造成的成像边缘存在模糊区域问题,提出一种基于非下采样轮廓波变换(Nonsubsampled Contourlet, NSCT)域平均梯度能量驱动的红外图像(IFR)与可见光图像(VBI)融合算法,首先对IFR和VBI分别进行NSCT变换,得到低频、中频、高频子带系数,低频融合采用平均梯度能量(Average Gradient Energy, AVGE)取最大的方法进行融合,中频融合采用空间频率(Spatial Frequency, STF)取最大值进行融合,高频提出了一种平均梯度能量驱动脉冲耦合神经网络(Average Gradient Energy Pulse Coupled Neural Network, AVGE-PCNN)的融合方法进行融合,采用逆非下采样轮廓波变换(Inverse Nonsubsampled Contourlet Transform, INSCT)得到最终融合图像,实验采用三组不同场景的IFR和VSI图像进行融合处理。通过对比实验证明,提出的融合方法在改善图像边缘模糊方面效果良好,主观评估和客观评价均优于DWT、DTCWT、NSCT算法。Aiming at the problem of blurred areas at the imaging edge caused by the fusion of infrared and visible light images,a fusion algorithm of infrared image(IFR)and visible light image(VBI)driven by the average gradient energy in NSCT domain was proposed.Firstly,NSCT transformation was performed on IFR and VBI respectively.The low frequency,intermediate frequency,and high frequency subband coefficients were obtained.The low frequency fusion adopts the method of taking the maximum average gradient energy(Average Gradient Energy,AVGE)for fusion,the intermediate frequency fusion adopts the maximum value of the spatial frequency(Spatial Frequency,STF)for fusion,and the high frequency A fusion method of Average Gradient Energy Pulse Coupled Neural Network(AVGE-PCNN)was proposed for fusion,and the final fusion was obtained by using Inverse Nonsubsampled Contourlet Transform(INSCT)In the experiment,three groups of IFR and VSI images of different scenes were used for fusion processing.The comparative experiments proved that the fusion method proposed in this paper has a good effect in improving image edge blur,and the subjective evaluation and objective evaluation are better than DWT,DTCWT,and NSCT algorithms.

关 键 词：红外图像 可见光图像 脉冲耦合神经网络 平均梯度能量 非下采样轮廓波变换 

分 类 号：TH138.5[机械工程—机械制造及自动化]