Effective hyperspectral image block compressed sensing using thress-dimensional wavelet transform

In this paper, an effective block compressed sensing algorithm based on improved noise variance estimation method is proposed for hyperspectral images. The reconstruction process adopts the iterative projected Landweber and soft-thresholding bivariate shrinkage image denoising based on three-dimensional wavelet transform. The improved noise variance estimation method can more effectively remove noise and achieve better image reconstruction quality. Experimental results demonstrate that the proposed algorithm significantly outperform several state-of-the-art compressed sensing algorithms.