The polarization detection technique provides polarization information of objects which conventional detection techniques are unable to obtain. In order to fully utilize of obtained polarization information, various polarization imagery fusion algorithms have been developed. In this research, we proposed a polarization image fusion algorithm based on the improved pulse coupled neural network (PCNN). The improved PCNN algorithm uses polarization parameter images to generate the fused polarization image with object details for polarization information analysis and uses the matching degree M as the fusion rule. The improved PCNN fused image is compared with fused images based on Laplacian pyramid (LP) algorithm, Wavelet algorithm and PCNN algorithm. Several performance indicators are introduced to evaluate the fused images. The comparison showed the presented algorithm yields image with much higher quality and preserves more detail information of the objects.
In order to achieve the high-resolution multispectral image, we proposed an algorithm for MS image and PAN image fusion based on NSCT and improved fusion rule. This method takes into account two aspects, the spectral similarity between fused image and the original MS image and enhancing the spatial resolution of the fused image. According to local spectral similarity between MS and PAN images, it can help to select high frequency detail coefficients from PAN image, which are injected into MS image then. Thus, spectral distortion is limited; the spatial resolution is enhanced. The experimental results demonstrate that the proposed fusion algorithm perform some improvements in integrating MS and PAN images.