一种改进的无链表3DSPITH算法用于高光谱图像压缩

Modified listless 3DSPITH with ROI for hyperspectral image compression

摘要: 针对大孔径静态干涉成像光谱仪(LASIS)的成像特点，提出了一种基于三维非对称等长树小波变换的无链表SPITH算法结合ROI的图像压缩方案。首先，对干涉高光谱图像进行三维非对称等长树离散小波变换。其次，采用ROI方法对主要的光谱系数进行保护。最后，采用改进的三维无链表SPITH算法，编码干涉高光谱图像的小波变换域。实验结果表明，该方法在8∶1压缩比下，获得大于40 dB的平均峰值信噪比，同时有效地保护了光谱信息。

Abstract: A compression algorithm based on asymmetric uniform-tree 3D-DWT and modified listless 3DSPIHT in conjunction with region of interesting (ROI) is presented. It is specially developed for the large aperture static imaging spectrometer (LASIS) system according to its interference properties. Firstly, the hyperspectral interference image series with asymmetric uniform-tree 3D-DWT are decomposed. Secondly, the major hyperspectral coefficients are protected by ROI method. Finally, a modified listless 3DSPIHT algorithm to encode the transformed images is given. The numerical experiment results show that the average PNSR is more than 40dB at 8:1 compression rate, and the efficient protection of hyperspectral information is achieved.