Sparse Representation of Injected Details for MRA-Based Pansharpening

Abstract

Pansharpening is a notable remote sensing topic in which high spatial resolution panchromatic image and low spatial resolution multi-spectral image are being fused in order to receive the high spatial resolution multi-spectral image. This paper presents a hybrid pansharpening method based on MRA framework and the sparse representation of injected details. To add spatial details of the panchromatic image into the multispectral image more effectively, the injection gains are computed through an iterative full-scale model in which the gains are updated at each iteration relying on its previous iteration's fusion product. The proposed method is compared with five pansharpening approaches to investigate the effectiveness. Experiments have been implemented on two data sets from the Pleiades and GeoEye-1 satellites both at reduced and full scale. In terms of visual and quantity assessment, the high-resolution MS image produced by the proposed method is more acceptable than those images fused by other rival approaches.