Elimination of unwanted anomalies in a hyperspectral image using modified subspace RX algorithm

Anomaly detection have been studied over few decades and successful results have been given in the literature. Local and global anomaly detection schemes are usually used separately and their strong points are not used perfectly. In this paper, we propose a new method both using local and global anomaly methods. Additionally, we propose a new modified local anomaly technique in between the dual-window Reed-Xiaoli (DWRX) and subspace RX (SSRX). The results show that using this new technique improves precision and recall values while the global anomaly method prevents the unwanted known objects to be detected as anomaly.