The automatic background extraction works according to the following algorithm:
The idea behind this algorithm is that background pixels are noisy pixels which do not exhibit intensities above or below a certain threshold (in none of the layers). So if a particular pixel does not "light up" (i.e. it falls outside the specified percentile range) in at least one of the layers this pixel does not contribute to the information content of the hyperspectral image, and is regarded as a background pixel. The thresholds for the percentile range are by default set to 5 and 95 percent.
- Initialize all pixels of a mask to TRUE
- For each layer
- Calculate the n- or the (100-n)-percentile of the numeric values in this layer
- Reset all mask pixels whose numeric value is below or above the n- or the (100-n)-percentile
- Background pixels are those pixels which remain in the TRUE state.