Adaptive hybrid pansharpening: a novel approach for combining two methods to achieve superior pansharpening performance

Abstract

ABSTRACT This study proposes an adaptive strategy to improve the performances of two widely used pansharpening strategies, namely the Generalized Laplacian Pyramid (GLP) with Modulation Transfer Function (MTF)-matched filter and Context-Based Decision (CBD) injection scheme (i.e. MTF-GLP-CBD) and the Sparse Representation of Injected Details (i.e. SR-D). By considering the variance information of the PAN data, the suggested method determines which part of a pansharpened image should be produced by which of these procedures. The effectiveness of the suggested technique was assessed both qualitatively and quantitatively by comparing its performance to 31 other pansharpening strategies across three different test sites with diverse features. The outcomes of the experiments demonstrated the efficacy of the suggested approach. The experiments showed that the suggested approach not only surpassed the standard MTF-GLP-CBD and SR-D algorithms, but also all the other pansharpening algorithms employed. Of all pansharpening results, the results of the suggested approach led to the best quality index values. It can also be concluded that the suggested technique has the potential to enhance the effectiveness of other pansharpening algorithms.