LAUSR

The spatial details injection model has been considered as a general framework in the literature of pansharpening, and recently there have been significant advances in this framework based on sparse representation (SR) of spatial details. However, the SR-based methods have greater computational burden in estimating the sparse vectors and limited ability in detail edge preservation. In this article, we introduce the neighbor embedding (NE) instead of SR-based model and the edge-preserving filter into the spatial detail injection framework to address the aforementioned two drawbacks. By utilizing the best quality of NE, we propose the detail injection via NE (DINE) algorithm for pansharpening, and DINE+, an improved variant of DINE by using the edge-preserving filter to enhance the spatial details. Experiments carried on three datasets captured by different satellite sensors and compared with current state-of-the-art methods validate the effectiveness of the proposed methods.