Anomaly Detection with Bayesian Gauss Background Model in Hyperspectral Images

Abstract

In this paper we propose Bayesian Gaussian Background Model (BGBM) for anomaly detection problem on hyperspectral images. BGBM is used in a local window for learning the background. BGBM gives better results than existing methods on real hyperspectral images in the estimation of covariance matrix in case of limited number of samples.