Fig. 9
From: Exposed soil and mineral map of the Australian continent revealing the land at its barest
Comparison of the effect of different weighting schemes on the output of the proposed algorithm. The effect of making choices of weight function \(\tilde w\) shown across an area to the west of Canberra, Australia. The location was chosen due to the presence of urban, water, vegetation and soil classes. The images are shown in false colour with SWIR1, NIR and BLUE in the RGB channels to accentuate healthy vegetation in bright green, soils in magenta and water in blue/black. The histogram stretches for all images are the same and chosen based on a 2–98% cutoff of the histogram of b which is an unweighted GM pixel composite mosaic. The weighting schemes are as follows: a is NDVI, c is −1 × NDVI, d is GNDVI, e is −1 × GNDVI, f is −3 × GNDVI, g is SAVI, h is −1 × SAVI and i is −3 × SAVI. We observe that c, f and i show the presence of the most soil across the image, however, f and i preference the presence of water (over soil) in some locations. NDVI provides good symmetry with respect to the GM case in b, in that a provides a highly vegetated composite and c a composite dominated by soil.
