Denoising hyperspectral images with non-white noise based on tensor decomposition

The hyperspectral images (HSIs) imply the exploration and the collection of a huge amount of data. Commonly, filtering methods for HSIs are based on the data vectorization or matricization while ignore the related information between image planes. So there are new approaches considering multidimensional data as whole entities, for example multidimensional Wiener filtering (MWF). However, it can not cope with the HSIs disturbed by non-white noise which is the most cases in the actual world. To remove non-white noise from images, a new method is proposed in this paper. It dose a prewhitening procedure to the original HSI to change the noise being a white one, then MWF can help to denoise the prewhitened data, in the end an inverse prewhitening processing is used to rebuilt the estimated signal. Comparative studies with other denoising methods show that our approach has promising prospects in this field.