基于三维整数小波变换的高光谱图像编码方法  

Hyperspectral image coding based on three-dimensional integer wavelet transform

作　　者：黄菁[1] 朱日宏[1] 李建欣[1] 

机构地区：[1]南京理工大学电光学院,南京210094

出　　处：《仪器仪表学报》2007年第12期2274-2279,共6页Chinese Journal of Scientific Instrument

基　　金：中国空间技术研究院创新基金(CAST200641);江苏省现代光学技术重点实验室开放课题(T8108030)资助项目

摘　　要：高光谱图像是一种数据量很大的三维图像,在存储和传输之前必须对其进行有效的压缩。本文提出了一种基于三维整数小波变换的高光谱图像压缩编码方法,首先利用三维整数小波包变换去除高光谱图像的空间和谱间冗余,然后对变换后的每个子带的小波系数构造子带重要性树,通过子带重要性树的分裂对小波系数进行三维嵌入零块编码,最后为了进一步提高编码性能,把编码输出的二进制符号送入算术编码器并结合高效的上下文模型进行熵编码。对4组场景的AVIRIS高光谱图像进行测试,结果表明该方法的编码性能略优于3DSPIHT和3DSPECK算法,同时该方法还支持高光谱图像从有损到无损的渐进传输。Hyperspectral images are massively large sized three-dimensional data sets. Therefore efficient compression should be applied to these data sets before storage and transmission. A three-dimensional wavelet-based coding algorithm is proposed for hyperspectral image compression. A three-dimensional integer wavelet packet transform is applied to reduce the spatial and spectral redundancies. Then the sub-band significance tree is built up for the wavelet coefficients of each wavelet sub-band, and three-dimensional embedded zero-block coding is performed on the wavelet coefficients via sub-band significance tree splitting. In order to further improve the performance, high order conditional arithmetic coding is used to encode the binary, symbols generated by the embedded zero-block coding. Experimental results on four AVIRIS images show that the proposed algorithm has a little higher performance compared with 3DSPECK and 3DSPIHT algorithms. Besides having high performance, this algorithm also supports progressive transmission of hyperspectral images.

关 键 词：图像编码 高光谱图像编码 嵌入零块编码 整数小波变换 

分 类 号：TN919.81[电子电信—通信与信息系统]