基于波段分组的3D-SPIHT高光谱图像无损压缩算法

Lossless Compression of Hyperspectral Image Based on 3D-SPIHT Using Band Classification

波段间隔为纳米级的高光谱图像具有很强的谱间相关性,但不同频谱波段图像之间的相关性不同,本文提出了一种基于波段分组的3DSPIHT(setpartitioninginhierarchicaltrees)高光谱图像无损压缩方法。对高光谱图像按照谱段类型进行分组,接着通过3维整型小波变换,对图像组去除空间相关性和光谱维相关性,最后以3DSPIHT的空间方向树组织方式来进行编码,去除小波变换后子带间系数的冗余。实验结果表明,该方法能够有效地去除空间和谱间相关性,在算法复杂度和计算时间上较整体处理有一定优势,同时可获得较好的无损压缩结果。

The spectral correlation in hyperspectral image is highly correlated, but it differs in different bands. A lossless compression scheme based on band grouping and 3D-SPIHT algorithm is proposed in this paper. First, we divide the image bands in different groups according to the band’s type. Then three dimensional integer wavelet transform is applied to each group, which can exploit both the spatial and spectral correlation. Finally, three-dimensional SPIHT algorithm is used to encode the tree-like wavelet coefficients. Experiments show that this algorithm can achieve better lossless compression as well as low complexity.