Pansharpening Based on Convolution Sparse Representation and NSCT

Abstract

In order to make full use of the spatial detail information of remote sensing images, a remote sensing image fusion method based on convolution sparse representation and non-subsampled contourlet transform (NSCT) was proposed. First, the convolution sparse representation is used to establish a model to complete the super-resolution of image and achieve the purpose of detail enhancement. Then, the two images are fused, and the super-resolution image and panchromatic image are subjected to NSCT transformation to obtain their respective high-resolution sub-band images and low-resolution sub-band images. Appropriate methods are adopted according to the characteristics of different sub-bands. The new sub-band information is obtained by the fusion rules, and finally the NSCT inverse transform is performed to obtain fusion result. Experiments proved that the fusion image obtained by this method is superior to these of other methods in both visual effects and objective indicators.

Keywords

• non-subsampled contourlet transform
• remote sensing image
• convolution sparse representation
• fusion rules