Multi-band infrared targets discrimination based on tensor filters

The correlativity between spatial and spectrum is not effectively used when the multi-band images were processed based on the traditional methods of vector. In the design of tensor discrimination method, the multispectral images as a whole was used to improve the discrimination performance of the traditional method. A Gabor tensor matched filter model was proposed which improved the traditional matched filters and utilized the spatial and spectral information. This method could make full use of the original spatial and spectral structures of the multispectral images, and improve the discrimination performance extremely efficient by using tensor model.