NSCT域内基于自适应PCNN的红外与可见光图像融合方法

提出一种在图像的非下采样Contourlet变换(NSCT)域内基于脉冲耦合神经网络(PCNN)的融合方法。首先采用NSCT对严格配准的待融合图像进行多分辨率多方向分解, 得到低频子带和高频方向子带；然后使用各子带系数的空间频率作为PCNN对应神经元的自适应连接强度系数，使用改进的拉普拉斯能量和作为PCNN每个神经元的外部激励，经过PCNN点火过程获得各子带对应的点火映射图，并通过判决选择算子确定融合图像的各子带系数；最后采用NSCT逆变换对低频子带系数和高频方向子带系数进行重构，得到融合图像。使用红外与可见光图像进行仿真实验的结果表明，本文方法优于基于小波变换、NSCT及传统NSCT与PCNN结合的图像融合方法。  

Fusion Algorithm of Infrared and Visible Images Based on Adaptive PCNN in NSCT Domain

In the nonsubsampled contourlet transform（ NSCT） domain,an image fusion algorithm based on pulse coupled neural networks（ PCNN） is proposed. Firstly,registered images to be fused are decomposed into low-pass subband and band-pass directional subbands by using NSCT. Secondly,sum-modified-Laplacian in NSCT domain is inputted to motivate PCNN whose linking strength is adaptive because of using spatial frequency in NSCT domain as the linking strength. Then,the sum of neuron firing times will generate a firing map,and coefficients of the fused image are selected from the coefficients in NSCT domain by the decision operator. Finally, the fused image is obtained by inverse NSCT. The results of simulation experiments with infrared and visible images show that the proposed algorithm has better performance than wavelet-based,NSCT-based,and typical NSCT-PCNN-based fusion algorithms.