A lossless compression scheme consisting of a linear prediction and multiband lookup tables was proposed to compress the airborne hyperspectral imagery efficiently. Firstly, based on the Yule-Walker equation, a linear prediction model whose equation coefficient matrix is a non-Toeplitz type covariance matrix and it should be solved by an extension form of Levinson algorithm was established by exploiting the strong correlation of spectral bands of hyperspectral imagery. Then, a multiband lookup table algorithm was adopted to refine the prediction result based on the calibrated hyperspectral imagery containing a sparse histogram induced by calibration techniques. However, for the uncalibrated imagery, the multiband lookup tables could be neglected. Finally, the prediction residuals were sent to the entropy encoder. In the experiment, the Adaptive Arithmetic Code and Golomb-Rice Code were both tested as the entropy encoder. The experimental results show that the proposed scheme has a higher compression ratio and the compression effect is better than that of the standard from Consultative Committee for Space Data System(CCSDS).