Abstract: Hyperspectral images have higher spectral resolution than normal images，however this advantage is at the cost of massy quantity of data that brings difficulties to image storage and transmission.So it is very important to search an effective compression algorithm for hyperspectral images.A hyperspectral image compression algorithm based on spectral prediction is proposed in this paper.Considering that as the spectral resolution of hyperspectral image increases，its spectral correlation becomes more significant，the algorithm used to solve the coefficients of the predictor by using multi-band to predict the current band linearly is derived，and an optimization way of selecting reference bands is proposed.The experiment results show that lower MMSE can be got，and it runs fast，so it works efficiently in practice.

Key words: hyperspectral images, interspectrum-prediction, compression, optimization

摘要： 高光谱图像具有较高谱分辨率的优越性是以其较大的数据量及较高的数据维为代价的，因此有必要研究有效的高光谱图像压缩方法。探讨一种基于谱间预测的高光谱图像压缩方案。考虑到高光谱图像谱间相关性随分辨率的提高而增强，推导出由多个波段对当前波段进行线性预测的预测器系数求解算法，提出了一种参考波段优化选取方法。实验结果表明，该方法能获得较低的最小均方误差，运算速度快，具有实用价值。

关键词: 高光谱图像, 谱间预测, 压缩, 优化