Abstract: A fast compression algorithm for hyperspectral images based on dispersion sorting in transform domain is proposedConsidering the characteristics of hyperspectral data in the Hadamard domain, the proposed algorithm selects a favourable order adaptively and sorts the dimensions of spectral vectors by dispersion. Consequently, the energy and difference of the spectral vectors is concentrated on the lower dimensions and the dimensions of high signal to noise ratio are moved into low dimensional subspace. Then, efficient eliminating inequalities are constructed. When combinined with the LBG(Linde Bazo Gray) clustering algorithm, the proposed algorithm quickly completes the encoding of hyperspectral images via vector quantization. Experiments were conducted under different compression ratios The results show that, the compression algorithm for hyperspectral images as presented in this paper can reduce the computational complexity significantly when completing fast compression based on the precondition of good recovery quality.