基于IPCNN的红外与可见光图像融合算法  

Infrared and visual image fusion algorithm based on IPCNN

作　　者：江泽涛[1,2] 吴辉[1] 周哓玲 黄锦 JIANG Ze-tao;WU Hui;ZHOU Xiao-ling;HUANG Jin(Key Laboratory of Image and Graphic Intelligent Processing of Higher Education in Guangxi,Guilin University of Electronic Technology,Guilin 541004,China;Guangxi Key Laboratory of Trusted Software Guilin University of Electronic Technology,Guilin 541004,China)

机构地区：[1]桂林电子科技大学广西高校图像图形智能处理重点实验室,广西桂林541004 [2]桂林电子科技大学广西可信软件重点实验室,广西桂林541004

出　　处：《计算机工程与设计》2018年第11期3475-3480,3493,共7页Computer Engineering and Design

基　　金：国家自然科学基金项目(61572147);广西科技计划基金项目(AC16380108;2015BC19022);桂林电子科技大学图像图形智能处理重点实验基金项目(GIIP201501);广西可信软件重点实验室基金项目(kx201502);广西研究生教育创新计划基金项目(YJCXS201536;2016YJCX71)

摘　　要：为充分保留源图像的细节信息,提高融合图像的清晰度、对比度,提出一种基于改进脉冲耦合神经网络(improved pulse coupled neural network,IPCNN)的红外与可见光图像融合算法。对源图像进行非降采样轮廓波变换(nonsubsampled contourlet transform,NSCT),采用基于静态小波变换(static wavelet transform,SWT)的融合策略对低频子带进行融合,对高频子带采用绝对值取大与IPCNN相结合的融合方式,在融合过程中引入链接突触计算神经网络(linking synaptic computation network,LSCN)进行图像增强,通过NSCT逆变换得到融合图像。实验结果表明,该算法的融合图像在清晰度、对比度、图像信息熵等方面均具有较好的优势。To fully preserve the detailed information of the source images and the clarity along with increasing the contrast of the fusion image,an infrared and visible image fusion algorithm based on improved pulse coupled neural network(IPCNN)was proposed.By decomposing the two original images into different frequencies with nonsubsampled contourlet transform(NSCT),the low frequency component was fused by the fusion strategy based on the static wavelet transform(SWT).For the high frequency components obtained by the NSCT,the combination of absolute value and IPCNN was applied to the high frequency components,linking synaptic computation network(LSCN)was introduced into the fusion process to enhance the image.The fusion image was obtained by performing the inverse NSCT on the fused components.Experimental results show that the fusion image has good superiority in clarity,contrast,image entropy and so on.

关 键 词：图像融合 脉冲耦合神经网络 非降采样轮廓波变换 静态小波变换 链接突触计算神经网络 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]