基于FPDEs与CBF的红外与可见光图像融合

针对传统红外与可见光图像融合结果中的对比度不足、块状效应、伪影以及边缘区域信息失真等问题,文中提出一种基于四阶偏微分方程(FPDEs)和交叉双边滤波器(CBF) 的红外与可见光图像融合方法。首先,分别使用FPDEs和CBF方法从源图像中获取近似层和细节层；其次,针对多尺度分解获得的近似层含有残余低频信息导致融合图像的整体视觉反差较大的问题,采用基于视觉显著性映射(VSM)的方法对近似层进行融合；然后,对细节层使用改进的Karhunen-Loeve变换获得权重,而后进行细节层融合；最后,通过线性组合方式将近似层与细节层融合,从而产生融合图像。实验结果表明,经基于FPDEs与CBF的方法融合后,相较于基于主成分分析和基于交叉双边滤波器的方法,基于FPDEs与CBF的方法所得融合图像的标准差 平均 提高了43.73%左右；相较于基于引导滤波器和基于视觉显著性最小二乘优化的方法,融合图像的平均梯度提高了约9.46%,空间频率平均提高了19.79%左右。

Infrared Image and Visible Image Fusion Based on FPDEs and CBF

Considering the problems of low contrast,blocky effects,artifacts and distortion of the edge region in traditional fused images,this paper proposed an infrared image and visible image fusion method based on fourth order partial differential equations(FPDEs) and cross bilateral filter(CBF).Firstly,the FPDEs and CBF are respectively used to obtain the approximation layers and detail layers from the source image.Secondly,the approximate layers obtained by multi-scale decompositions may contain amount of residual low-frequence information which will result in large contrast of the overall visual of the fused image,so a fusion method based on visual saliency map(VSM) is used to fuse the approximate layers.Thirdly,an improved Karhunen-Loeve transform is applied into the detail layer to obtain the optimal weights for fusion.Finally,a fused image is generated from the linear combination of final approximate layers and detail layers.Experimental results show that the standard deviation of the fused image obtained by the proposed method increases about 43.3% than PCA based method and cross bilateral filter based method,and the average gradient and spatial frequency increase about 9.46% and 19.79% respectively on average compared with GFF and VSM_WLS algorithms.