Anomaly detection for replacement model in hyperspectral imaging

In this paper we consider Anomaly Detection in the hyperspectral context, and extend popular RX detector, initially designed under standard additive model, to replacement model case. Indeed, more realistic framework, target, if present, is supposed replace a part of background. We show how estimate background power variation improve scheme. The obtained Replacement (RRX) shown be closed-form outperforms on real data benchmark experiment.