A NSST Pansharpening method based on directional neighborhood correlation and tree structure matching

In this paper, we propose a multispectral (MS) remote sensing image pansharpening method based on non-subsampled shearlet transform (NSST). By analyzing the NSST high-frequency coefficients correlation of several datasets which are fromWorldView-2 (WV2) and Quick-Bird (QB), we verified that the high-frequency coefficients based on NSST have strong directional neighborhood correlation within the same sub-band and parent-children correlation between sub-bands in the same direction. In order to combine these two kinds of correlations, we design a type of weighted directional neighborhood templates which can be used for any number of direction sub-bands to depict the direction correlation, and use the tree structure to model the correlation between parent-children coefficients. Experiments show that the proposed method in this paper can provide a fused MS image with high spatial resolution, which can provide convenience for subsequent applications such as classification and target recognition.