非下采样 Contourlet 变换和脉冲耦合神经网络相结合的遥感图像融合方法

摘要

In order to optimally preserve the spectral information of multi-spectral images and maximally fuse the spatial information of panchromatic images,we present an image fusion method by combining the nonsubsampled Contourlet transform (NSCT)and pulse coupled neural networks (PCNN)in this paper.First,we divide target area into edge area and non-edge area according to spatial distribution characteristics of ground objects.We employ spatial domain fusion with panchromatic image and I component of multispectral image in non-edge area to improve spectral information.Second,spatial fused image and I component of multispectral image are transformed by NSCT,and we regard local energy and spatial frequency as original information of source image to motivate PCNN and fuse image by taking large number of ignition as fused rules.Experimental results demonstrate that the proposed method is feasible in both keeping spectral information at non-edge area and improving spatial information at edge area.%为最优保留多光谱图像光谱信息的同时，最大限度地融入全色图像的高空间信息，该文提出了一种基于非下采样 Contourlet（非自适应方向多尺度分析方法）变换和脉冲耦合神经网络相结合的图像融合的方法。根据目标融合区域地物的空间分布特点，将目标融合区域划分为边缘区域和非边缘区域，并对全色图像和多光谱图像 I 分量在非边缘区域进行空间域融合，融入更多多光谱图像的光谱信息。然后，对多光谱图像 I 分量和空间域融合后的图像进行非下采样 Contourlet 变换，在低频子带和高频子带分别采用区域能量和空间频率作为源图像的原始信息，驱动脉冲耦合神经网络以每个像元的点火数作为活跃性测量，对图像进行融合。实验结果表明：该算法在非边缘区很好地保持了多光谱图像的光谱信息，在边缘区融入了更多的全色图像的空间细节信息，提高了融合图像的空间分辨率。