A solution to the problem of automated detection of targets with unknown spectral properties in multispectral imagery is presented that makes use of three background characterization and suppression algorithms in series. The first, parametric Bayesian clustering, is used to accurately characterize individual elements of the background scene. The second, background suppression filtering, eliminates those dimensions of multispectral space containing the majority of background energy. Finally, a multidimensional extension of the well-known Linde- Buzo-Gray (LBG) clustering algorithm is used to characterize what remains of the background and extract any anomalous target signatures. The results of this process are compared to spectral decorrelation (RX) filtering alone, LBG clustering alone, and RX filtering in combination with background suppression filtering. The process presented is shown to be significantly superior to each of these algorithm combinations.