超光谱图像是三维立体图像，具有较高的光谱分辨率和较多的光谱通道数。但较高的光谱分辨率是以较大的数据量和较高的数据维为代价的。庞大的数据量给存储和传输带来了一定困难，因此对超光谱图像进行压缩是非常必要的。由于超光谱图像是一种重要的数据源，特别要求压缩算法实时性好、可靠性高，因此应尽可能采用无损或近无损压缩方法。本文对超光谱遥感图像的压缩算法进行研究。 超光谱图像具有较强的空间相关性和很强的谱间相关性，针对超光谱图像的这一特性，本文采用了基于三维小波变换的3D-SPIHT编码算法。针对超光谱图像的特性，采用三维小波变换，同时去除空间冗余和谱间冗余。然后根据变换后小波系数的特性，构造一种3维空间方向树结构，用3D-SPIHT算法对小波系数进行量化编码。 在对图像进行小波变换时，本文选择了两种小波基：9/7小波和5/3小波。仿真结果表明，在传输速率为1.0bpp时，9/7小波的峰值信噪比为41.557db；而5/3小波仅为36.350db，相比9/7小波下降了12.5%。在对超光谱图像进行去除谱间冗余时，本文采用了基于DPCM谱间去相关和基于小波变换的谱间去相关两种方案。实验数据表明，采用这两种方案对超光谱图像进行压缩，在压缩比特率为1.0bpp时，两种方案的峰值信噪比都达到了40db。但两种方案相比来说，9/7小波去除谱间相关性效果更好些。这说明基于9/7小波的3D-SPIHT编码更能适合超光谱图像的压缩。

Hyper-spectral image is a three-dimensional image, has the high spatial resolution and a large number of spectral channels. But the high spatial resolution is taken for the greater data amount and bigger dimension. The huge data amount has brought certain difficulty for the storage and the transmission, so compression to the image is necessary. Because the hyper-spectral image is a kind of important data, and it need fast speed and high reliability.Therefore lossless or near lossless compression method should be used as far as possible.In this paper, algorithms of hyper-spectral image compression are proposed. The Hyper-spectral image has much strong spatial correlation and spectroscopy correlation. According to these, 3D-SPIHT encoding algorithm,which is based on the three dimensional wavelet transformation(3D-DWT),has been used in the paper.And 3D-DWT was used to wipe off the redundancy of spatial and spectroscopy. Then a kind of 3 dimensional Spatial Orientation Tree was constructed by the coefficients of wavelet transform.Finally,the coefficients of wavelet transform were coded by 3D-SPIHT algorithm. Two kinds of wavelet,9/7 wavelet and 5/3 wavelet,were used in the paper.They are both base on the wavelet transformation. Simulation results show ,when transmission bit rate is 1.0bpp,PSNR of 9/7 wavelet is 41.557db,PSNR of 5/3 wavelet is 36.50db.Compares with 9/7wavelet,PSNR of 5/3 wavelet drops 12.5%.The paper adopts two programs on removing spectroscopy correlation of hyper-spectral images:DPCM spectral compression and wavelet transform spectral compression. The experimental data shows,using the two programs on compression of hyper-spectral , PSNR can be both achieved above 40db, when they both at 1.0bpp transmission bit rate.But compared,9/7 wavelet has better effect on removing spectroscopy correlation.PSNR enhance about 1db. This show that the progranm which based on 9/7 wavelet 3D-SPIHT coding algorithm is much more adapted to the hyper-spectral image compression.