A spectral-preserving pansharpening method for multisensor satellite imagery

Abstract

Pan-sharpening is a data fusion process that combines lower resolution multispectral (MS) and high-resolution panchromatic (PAN) imageries to produce a high-resolution color image. For the commonly used pansharpening algorithms the intensity of MS, usually derived from combining digital numbers (DN) of each band, is replaced with the intensity of PAN. Thus, the spectral content of the pan-sharpened image is not the same as that of MS. In this paper, we implement a simple approach of pan-sharpening which produces pan-sharpened images with high spatial resolutions as in PAN while maintaining the spectral fidelity of the MS reflectance. The DN of MS is converted to the ground reflectance. The overlap of the spectral response function (SRF) of each MS band with the PAN SRF is used as the relative weight of the respective band to calculate the effective MS intensity. A look-up-table (LUT) of the effective MS intensity to PAN DN is derived. Finally, the resolution of the MS reflectance is enhanced by the ratio of PAN to the LUT value of the effective MS intensity. Since the DN of PAN and LUT value of the effective MS intensity have the same mean, the mean of the resolution-enhanced MS reflectance remains the same. The method has been applied successfully to sharpen the MS images of Sentinel-2 with PAN images of TeLEOS-1 satellites.