基于3D SPIHT的高光谱图像压缩技术

将三维多级树集合分裂(3D SPIHT)算法用于高光谱图像的压缩。根据高光谱图像的特点进行波段分组以得到处理单元,对各组分别进行三维小波变换,去除谱间和谱内冗余;利用3D SPIHT算法对变换后小波系数进行编码,去除系数之间的冗余。采用整数小波和浮点小波分别进行无损和有损压缩仿真,AVIRIS和OMIS实验结果表明,在bit/pixel为1的条件下,平均PSNR比准三维方法分别高0.91 dB和1.38 dB,且算法具有嵌入式、可伸缩性等优点,但无损压缩的平均bit/pixel比基于预测的方法高0.308和0.159。3D SPIHT算法用于高光谱图像压缩可以获得较好的有损性能,但无损压缩性能逊于基于预测的方法。

Hyperspectral image compression based on 3D SPIHT algorithm

A 3D SPIHT algorithm is applied to hyperspectral image compression.The band is separated into several groups to obtain the processing unit,and then 3D wavelet transfor(WT) is applied to each group to eliminate both spatial and spectral redundancies.Finally,the 3D SPIHT algorithm is used to encode the wavelet coefficients.Both integer and float wavelets are used for lossless and lossy compression respectively.The experimental results on AVIRIS and OMIS indicate that the average PSNR by proposed algorithm can be 0.91 dB and 1.38 dB higher than that of general 3D WT coding method for lossy compression,but 0.308 bit/pixel and 0.159 bit/pixel higher than the predictive coding technique for lossless compression,respectively.The proposed algorithm has better lossy coding performance,but lossless coding performance is inferior to that of predictive coding technique.