A New Fusion Algorithm for Optical Remote Sensing Data

Abstract

The main goal of data fusion in remote sensing is integrating the respective superiority and complementary information of multi-source data to serve the applications. Most conventional fusion algorithm is oriented to fuse the multispectral (hyperspectral) data with panchromatic data such as IHS fusion, PCA fusion, DWT fusion, Brovey fusion etc. Although these methods have been widely used, they still have a problem that can’t maintain the spatial information and spectral information synchronously. In this study, a new fusion algorithm aiming at fusing multispectral data and hyperspectral data was developed in following steps: First, classify the hyperspectral data into groups basing on the wavelength range of each multispectral band; second, process each band group by PCA, and combine all the first components to a new multispectral data; then, separate each original multispectral pixel by the homologous pixel and its eight neighbor pixels in the new multispectral data; Finally, mix the homologous 9 pixels of hyperspectral data by the abundance got by former step. The applying on AVIRIS data ‘cuprite’ brought by ENVI shows an excellent result and verified the validity.