Anomaly subspace detection based on a multi-scale Markov random field model

We introduce a multi-scale Gaussian Markov random field (GMRF) model and a corresponding anomaly subspace detection algorithm. The proposed model is based on a multiscale wavelet representation of the image, independent components analysis (ICA), and modeling each independent component as a GMRF. The anomaly detection is subsequently carried out by applying a matched subspace detector (MSD) to the innovations process of the GMRF, incorporating a priori information about the targets. The robustness of the proposed approach is demonstrated with application to automatic detection of airplanes on synthetic cloudy sky backgrounds.