Multiscale decomposition-based anomaly detection for hyperspectral images

An anomaly detection algorithm for hyperspectral images based on multiscale decomposition is proposed. Both spatial and spectral information is used to locate and detect targets under the condition of no prior knowledge about target and background. Firstly, the hyperspectral images are decomposed into a series of different scaled sub-bands using nonsubsampled pyramid decomposition. Then using the correlation of neighborhood coefficient of different scaled space in a hyperspectral band, the background data is optimally predicted by reducing the anomalous data using unsharped masking filter in different scale of each band and finally the anomaly targets can be detected by using designed kernel RX operator in the feature space. Numerical experiments have been conducted on real and synthesized AVIRIS data to validate the effectiveness of the proposed algorithm. Compared with classical RX algorithm, the proposed algorithm has better detection performance and lower false alarm probability.