Denoising of hyperspectral images using shearlet transform and fully constrained least squares unmixing

In this paper, we propose a new denoising method based on a 2D non-subsampled shearlet transform (NSST) and fully constrained least squares unmixing (FCLSU). In the proposed method, first low noisy (LN) bands are separated from high noisy (HN) bands using spectral correlation. Second, NSST is applied to each spectral band of the hyperspectral images. Third, LN bands are denoised using a thresholding technique on the shearlet coefficients and HN bands are denoised by applying FCLSU. The proposed method is compared to state of the art denoising methods on synthetic and real hyperspectral datasets. The effect of denoising on classification accuracy is also investigated. Obtained results show the superiority of the proposed approach.