Texture segmentation based anomaly detection in remote sensing images

Abstract

Anomaly detection is of great importance both in civil and military applications. To avoid the interference of complex background, we propose a texture segmentation based anomaly detection algorithm (TBAD). TBAD introduces the Gaussian Markov random field (GMRF) to model the distribution of background pixel values. Through the GMRF model, images are segmented into series of textures. And the conventional RX algorithm is applied on each segmented textures. Since TBAD can utilize both the spatial and the spectrum information of background, it can reduce the interference of complex background. Experiment applied to real image has validated the performance of the new approach.