High Level Description of the BandMax Algorithm
This is a high level description of the BandMax Algorithm which is part of ENVI's Target Finding with SAM and BandMax tool. Although the algorithm is proprietary, we can provide these details on how the algorithm works.
The BandMax algorithm determines separability between bands of target and background spectra. Here, separability refers to the amount of difference between the spectra in each set of bands. The separability is measured by comparing the ratio of reflectance between pairs of bands for the selected target and background spectra. This separability value for a pair of bands in one pair of target and background spectra is the difference between the ratio of the reflectance values in that pair of bands in the target spectrum and the ratio of the reflectance values for the same pair of bands in the background spectrum.
This value is calculated for each pair of bands in the input spectra, and the largest difference for each band defines the significance value for that band. A higher significance value implies greater separability between the target and background spectra for the specified band and indicates that a band more likely to distinguish between the input target and background spectra. A threshold between 0 and 1 can then be selected and used to define every band with a significance value greater than that threshold's "significant band". This set of significant bands can then be used (instead of all the bands) in an image analysis or mapping process to find instances of the target, while reducing the probability of detecting the specified background material.
For additional information on BandMax or how to run the tool, please refer to the ENVI help.
Review on 12/31/2013 MM