A new hyperspectral image compression method is proposed by use of the separable non-uniform decoupled 3-D wavelet transform. The method is designed according to the fact that the properties of the different dimensions are distinctly unlike in the hyperspectral images. Traditional 3-D wavelet transform uses same wavelet base for the different dimensions and the frequency spectrum of 3D filters is identical. So we apply different wavelet bases to different dimensions, and design the method with the decoupled wavelet structure. In order to obtain the optimal combination of the wavelet bases, four evaluation items are used. Result shows that the new 3-D hyperspectral image compression method can significantly improve the compression ratio without sacrificing other features such as scalability and computational efficiency.