Unsupervised hyperspectral target detection based on multiresolution image fusion

In this paper, a new unsupervised target detection method based on multiresolution image fusion is proposed for hyperspectral images. The proposed method mainly includes two key procedures: hyperspectral image fusion and unsupervised target detection. In the hyperspectral image fusion, wavelet-based multiresolution analysis is used to decompose and reconstruct images, and local variance in wavelet domain is adopted as fusion features. Automatic subspace decomposition (ASD) is first performed on original data before wavelet-based fusion. RX algorithm, which is classical and effective to unsupervised target detection, is used in the proposed method. The numerical experiments are conducted on AVIRIS data with 126 bands. The experiments results show that the proposed method is very effective to anomaly detection in hyperspectral images.