High resolution satellite imagery was simulated using airborne scanner imagery and two different fusion algorithms were applied to the data. The advantages of the fusion by relative contributions compared with HSV transformation are a better correlation between the original (1 m) and the fused (1 m + 4 m) imagery and the applicability to multispectral imagery with more than three bands. If the four times coarser multispectral image is smoothed before the fusion process, the resulting images do not only look better to the eye but also show a better correlation.
The comparison of fused and full resolution images showed that NDVI-computations and land cover classification strongly depend on the applied fusion method and scene content.
If fusion by relative contributions is used, the agreement of classification (kappa coefficient) between original and fused images is generally very high. The kappa coefficient decreases with an increasing number of classes, since more transitions between classes are introduced, where misclassification can occur due to the blurred spectral information. It emerged that the classification agreement also strongly depends on the variability of the scene content, i.e. the kappa coefficient is higher for images with a wider range of spectral classes.