A 3D SPIHT algorithm is proposed to deal with the compression of hyperspectral images, In order to eliminate the intraband and interband dependencies, 3D discrete wavelet transform is used to decorrelate the hyperspectral images. After transform, Said and Pearlman＇s SPIHT algorithm is used to code the wavelet coefficients. The spatial orientation trees of the wavelet coefficients are three-dimensional, so a 3D SPIHT algorithm is constructed to suitably compress the hyperspectral images. The test results show that the 3D SPIHT algorithm based on 3D DWT for hyperspectral images compression is very efficient, the algorithm is embedded and has modest complexity