基于双通道PCNN的NSST域红外与可见光图像融合

A Novel Infrared and Visible Image Fusion Algorithm Based on Dual-channel Pulse Coupled Neural Network in NSST Domain

针对传统红外与可见光图像融合中存在的一些不足,提出一种新的基于非下采样剪切波变换(NSST)和双通道脉冲耦合神经网络模型(2APCNN)的红外与可见光图像融合算法.该算法首先对红外图像进行预处理,提高源图像的对比度,再对红外与可见光图像进行NSST分解得到低频和高频子带系数;然后对分解后的低频子带系数进行二维小波分解再次得到相应的低频和高频子带,低频部分采用一种基于显著图的融合策略,高频部分采用绝对值取大的原则,之后再对低频和高频采用小波逆变换得到NSST重构所需的低频部分;接着对NSST分解后的高频子带采用双通道PCNN进行处理;最后对处理过的低频和高频子带进行NSST逆变换得到最终的融合图像.几组图像的实验结果对比显示,该算法相比其他算法在客观评价指标和视觉效果上均取得了一定的改进.

This paper proposes a novel infrared and visible image fusion algorithm based on nonsubsampled shearlet transform( NSST) and dual-channel pulse coupled neural network( 2APCNN) in order to overcome the shortages of traditional image fusion algorithm. Firstly,an S-function is used to adaptively enhance the contrast of the infrared image. Secondly,the infrared and visible images are decomposed into low frequency and high frequency sub-bands by NSST transform. Thirdly,the low frequency and high frequency sub-bands of the infrared and visible images are obtained from the obtained low frequency sub-band by using wavelet transform( DWT).The significant figure rule is employed to fuse the acquired low frequency sub-bands,and the maximum method is projected to fuse the acquired high frequency sub-bands. Then,the low frequency sub-bands of NSST reconstruction are obtained by inversion DWT. Fourthly,the adaptive 2APCNN is projected to fuse the high frequency sub-bands in NSST domain. Finally,the fused image is obtained by performing the inverse NSST. The experiment results show that the proposed approach can obtain state-of-the-art performance compared with other image fusion methods in term of objective evaluation criteria and visual quality.