干旱区生态环境知识资源中心

Image data sharpening is a challenging field of remote sensing science, which has become more relevant as high spatial-resolution satellites and superspectral sensors have emerged. Although the spectral property is crucial for mineral mapping, spatial resolution is also important as it allows targeted minerals/rocks to be identified/interpreted in a spatial context. Therefore, improving the spatial context, while keeping the spectral property provided by the superspectral sensor, would bring great benefits for geological/mineralogical mapping especially in arid environments. In this paper, a new concept was tested using superspectral data (ASTER) and high spatial-resolution panchromatic data (WorldView-2) for image fusion. A modified Principal Component Analysis (PCA)-based sharpening method, which implements a histogram matching workflow that takes into account the real distribution of values, was employed to test whether the substitution of Principal Components (PC1-PC4) can bring a fused image which is spectrally more accurate. The new approach was compared to those most widely used-PCA sharpening and Gram-Schmidt sharpening (GS), both available in ENVI software (Version 5.2 and lower) as well as to the standard approach-sharpening Landsat 8 multispectral bands (MUL) using its own panchromatic (PAN) band. The visual assessment and the spectral quality indicators proved that the spectral performance of the proposed sharpening approach employing PC1 and PC2 improve the performance of the PCA algorithm, moreover, comparable or better results are achieved compared to the GS method. It was shown that, when using the PC1, the visible-near infrared (VNIR) part of the spectrum was preserved better, however, if the PC2 was used, the short-wave infrared (SWIR) part was preserved better. Furthermore, this approach improved the output spectral quality when fusing image data from different sensors (e.g., ASTER and WorldView-2) while keeping the proper albedo scaling when substituting the second PC.