With the development of high spectral and spatial resolution imaging spectrometer,the amount of hyperspectral image data dramatically increase.Due to the limitation of transmission and storage ability,hyperspectral image must be compressed effectively.Firstly,the correlation of hyperspectral image is deeply analyzed,and it shows that the image has a certain spatial correlation and an extremely high spectral correlation.That makes it possible to compress the image to a great extent.Secondly,the scheme of lossless compression based on one-order linear prediction and JPEG2000 is proposed after the modification of DPCM combining with JPEG2000.Finally,the scheme is realized by programming and preferable results are obtained.The results show that hyperspectral image can be effectively compressed with the scheme,which verifies its feasibility and provides the theoretical basis for its implementation on the hardware platform.