Skip to Main Content
This study investigated the performances of data fusion algorithms when applied to very high spatial resolution satellite images that encompass ongoing- and post-crisis scenes. The evaluation entailed twelve fusion algorithms. The candidate algorithms were applied to GeoEye-1 satellite images taken over three different geographical settings representing natural and anthropogenic crises that had occurred in the recent past: earthquake-damaged sites in Haiti, flood-impacted sites in Pakistan, and armed-conflicted areas in Sri Lanka. Fused images were assessed subjectively and objectively. Spectral quality metrics included correlation coefficient, peak signal-to-noise ratio index, mean structural similarity index, spectral angle mapper, and relative dimensionless global error in synthesis. The spatial integrity of fused images was assessed using Canny edge correspondence and high-pass correlation coefficient. Under each metric, fusion methods were ranked and best competitors were identified. In this study, The Ehlers fusion, wavelet principle component analysis (WV-PCA) fusion, and the high-pass filter fusion algorithms reported the best values for the majority of spectral quality indices. Under spatial metrics, the University of New Brunswick and Gram-Schmidt fusion algorithms reported the optimum values. The color normalization sharpening and subtractive resolution merge algorithms exhibited the highest spectral distortions where as the WV-PCA algorithm showed the weakest spatial improvement. In conclusion, this study recommends the University of New Brunswick algorithm if visual image interpretation is involved, whereas the high-pass filter fusion is recommended if semi- or fully-automated feature extraction is involved, for pansharpening VHSR satellite images of on-going and post crisis sites.