Spectral unmixing

Spectral unmixing is the process of estimating for each image location the combination of elements of a set of endmember spectra that optimally approximates the image spectra at those locations. In linear spectra unmixing, the linear combination of endmember spectra that is optimal in the least squares sense is computed.

Spectral unmixing is most useful when most image of the image pixels represent a mosaic of multiple spectrally distinct material classes of interest. This is often the case when the spatial resolution of the image is low compared to the size of the objects of interest in the scene.

ENDMEMB is used to compute a set of endmember spectra using the image as input. These automatically computed endmember spectra may correspond to elementary materials in the imaged scene (e.g., various types of exposed soil, various agricultural crops, asphalt, concrete), but there is no guarantee that this will be the case. A comparison of the endmember spectra and reference spectra known to represent certain elementary materials is required to determine whether any endmember spectra represent elementary materials.

SPUNMIX performs linear spectral unmixing. The output of SPUNMIX is one map for each endmember spectra. The value at a given location in such a map is the estimated fractional contribution of the map's associated reference spectrum to the image spectrum at that location.

ENDMEMB and SPUNMIX are used when image variations that were not caused by terrain-cover variations have been removed from the image data. Typically this means that the image data have been corrected to scene radiance (for flat terrain), reflectance, or reflectance ratios.