Related to multivariate statistical analysis. Spectral unmixing is a procedure applied to images in which each pixel is associated with a measured spectrum, for example a spectrum of absorption for different wavelengths of light. In the spectral unmixing procedure, this measured spectrum is decomposed into a collection of constituent spectra, due to individual endmembers, such as different absorbers, and a set of corresponding fractions, or abundances, that indicate the proportion of each endmember present in the pixel.
https://www.ll.mit.edu/publications/journal/pdf/vol14_no1/14_1survey.pdf