A lossless compression algorithm of hyperspectral images based on two estimated value and look-up-table (LUT)prediction is proposed. Firstly,the intraband is predicted by using a JPEG-LS median predictor only for the first image along the spectral line,and the interband is predicted for the other bands. For interband prediction,the first estimated value is gotten by using three LUT prediction values,and then the second estimated value is computed by using eight specific pixels which located in the previous band and in the current band,the intraband prediction is effectively combined with the interband prediction,the spectrum correlation of hyperspectral images is removed. Then,the final estimated value is gotten through comparing the three LUT prediction values and the two estimated values. Lastly,the prediction error will be compressed by arithmetic coding. The experiment results show that:compared with LAIS-LUT,the proposed algorithm could enhance the compression ratio by about 0.03~0.11 for AVIRIS hyperspectral images data,and 0.01~0.09 for OIMS-I hyperspectral images,which improves the compression ratio effectively.