Quality assessment of ten fusion techniques applied on Worldview-2

Abstract

Fusion or merging or pansharpening is the digital processing that combines the high spatial and multispectral information to obtain a fused multispectral image that retains the spatial information from the high resolution panchromatic image, as well as the spectral characteristics of the lower resolution multispectral image. Many fusion algorithms were developed and tested on different commercial satellite data such as Ikonos and Quickbird, however, the efficacy of these algorithms is weakly assessed on the new satellite images of Worldview-2 with very high spatial resolution. In the present study ten well known fusion techniques namely Color Normalized (CN), Ehlers, Gram-Schmidt, High Pass Filter (HPF), Local Mean Matching (LMM), Local Mean and Variance Matching (LMVM), Modified IHS (Modihs), Pansharp, PCA and Wavelet are used for the fusion of Worldview-2 panchromatic and multispectral data. Fused images were evaluated for spectral and spatial fidelity using visual inspection and different quality indexes such as the correlation coefficient (CC), ERGAS, the Universal Image Quality Index (Q), the Expanded Quality Index (Q4_expand) and the Entropy values. Under each quality metric the fusion algorithms were ranked and best performances were indentified. It has been proved that all fusion algorithms have a different performance with the new multispectral bands of Worldview-2 in comparison to the classical ones. It has also been proved that almost all quality indices present a very strong dependency to the spatial resolution of the input data. The Q4_expand quality index is proposed as the main quality discriminator since it gave the most robust results independently of the spectral and spatial input resolution. Pansharp algorithm took the lead in spatial enhancement while Ehlers and HPF are ranked in the first places for spectral preservation.