基于NSST和自适应PCNN的图像融合算法

摘要

针对红外和可见光图像的特点，本文提出了一种基于非下采样剪切波变换（NSST）和自适应的脉冲耦合神经网络（PCNN ）相结合的红外与可见光图像融合的新算法。对经过NSST变换后的低频子带系数采用带高斯权重分布矩阵的局域方差和方差匹配度相结合的融合规则，对高频子带系数采用一种改进的空间频率作为PCNN输入，且采用改进的拉普拉斯能量和作为PCNN的链接强度，利用PCNN全局耦合性和脉冲同步性选择高频子带系数，最后经NSST逆变换后得到融合结果。实验结果表明，本文提出的算法与传统的图像融合算法相比不仅在主观视觉上取得较好的效果，而且在客观标准上也有了一定的提高。%Aiming at the feature of infrared and vision images,a new fusion algorithm which combines nonsubsampled shearlet transform (NSST)with adaptive pulse coupled neural network (PCNN )is presented.For the low-frequency sub-band coefficients,a fusion rule which combines local variance with a Gaussian weight distribution matrix after NSST transform with variance matching is used.For the high-frequency sub-band coefficients,an improved spatial fre-quency as the input of the PCNN is used,and the improved sum of Laplace energy as the PCNN link strength is used.The high-frequency sub-band coefficients are selected by using the global coupling and pulse synchronization of PCNN,and finally fusion results are obtained by inverse NSST transform.The experiment results show that compared to the traditional image fusion algorithms,the proposed algorithm achieves better results in the subjective visual and also improves the objective criteria in some extent.