基于非抽样剪切波变换的遥感图像融合方法

提出了一种基于Mean-shift分割和非抽样剪切波变换(NSST)的多光谱与高分辨率全色图像融合方法。对高分辨率图像进行Mean-shift分割，并利用区域方差将多光谱图像划分为需要进行空间细节增强及需要光谱特征保持的区域；然后利用NSST变换对高分辨率图像和多光谱图像的强度分量进行多尺度分解。分解后的低频子带采用基于四阶相关系数的融合规则进行融合，带通方向子带根据分割所得的区域按区域方差进行融合；最后进行NSST重构得到融合后的强度分量，经IHS逆变换获得高分辨率的多光谱图像。仿真实验表明，与其他4种相关的融合方法相比，该方法能在空间分辨率的提高与光谱信息的保持之间达到良好的平衡，使得融合图像不仅具有较好的光谱保持特性，而且其空间分辨率也能得到有效提高。

Fusion Method in Remote Sensing Image Based on NSST

A multispectral and panchromatic image fusion algorithm based on mean-shift segmentation and the non-subsampled shearlet transform (NSST) was presented. Mean-shift segmentation was performed on the panchromatic image, and the variance of each region was used to distinguish the multispectral image into regions that need to be spatially enhanced or not. Then, the multi-scale NSST was performed on the panchromatic image and the intensity component of the multispectral image. The low frequency components were fused based on the fourth-order correlation coefficient, and the band-pass frequency components were fused based on the variances of the segmented regions. At last, the fused intensity component was obtained by reconstruction of those fused components, and the inverse IHS transform was performed to obtain the fused high resolution multispectral image. Experimental results indicated that the proposed image fusion method could keep a balance between spectral preservation and spatial enhancement.