Anomaly Detection Method Based on Random Field for Hyperspectral Imagery
一种基于随机场模型的高光谱影像目标探测算法

This paper presented an anomaly detection method based on random field model in order to introduce the spatial information between the neighborhood pixels in the hyperspectral imagery into the anomaly detection procedure and reduce the area for detection. In our method, the pixels' neighborhood relationship in the hyperspectral imagery was described by the Random Field model. I}hen this neighborhood relationship information between pixels was introduced into the local-region anomaly detector which uses a nested dual window to detect probable anomaly pixels. Experiments show that this method performs better than the traditional RX-algorithm, especially for the larger anomaly targets which usually contains several neighborhood pixels and with better efficiency.