复数SAR图像的块自适应多级矢量量化压缩  

Complex SAR image compression using block adaptive multi-stage vector quantization

作　　者：喻言[1,2] 王贞松[1] 

机构地区：[1]中国科学院计算技术研究所,北京100190 [2]中国科学院大学,北京100049

出　　处：《高技术通讯》2014年第4期347-354,共8页Chinese High Technology Letters

基　　金：总装备部预研资助项目

摘　　要：研究了在空间域进行复数合成孔径雷达(SAR)图像的压缩,提出了结合块自适应矢量量化(BAVQ)和多级矢量量化(MSVQ)技术各自优点的块自适应多级矢量量化(BAMSVQ)压缩算法。该算法因SAR复数图像的实部和虚部数据在一定采样规模下具有近似高斯分布的特性,可以运用BAVQ编码;而上一级编码后的残差数据由于已经进行了一定程度的去相关性,根据中心极限定律也具有近似高斯分布特性,因而也可进行块自适应的矢量量化编码,由此可形成BAVQ的多级方案。另外由于实虚部归一化的高斯分布块数据具有相同的分布特性,因此可以共享码书,节省存储资源与传输数据量。仿真结果验证了该算法的有效性。在相同的压缩比下,与前两种空间域压缩算法相比,该算法提升了性能,降低了计算复杂度,节省了计算时间和资源占用量。A study on complex SAR image compression in the spatial domain was conducted,and an image compression algorithm based on block adaptive multi-stage vector quantization （BAMSVQ） was proposed.The BAMSVQ algorithm has the advantages of the block adaptive vector quantization （BAVQ） technique and the multi-stage vector quantization （MSVQ） technique.Based on the near Gaussian distribution of both the real part and the imaginary part of SAR images in a suitable block scale,the algorithm can use the BAVQ.The residual image data after the previous vector quantization（VQ） encoding stage also has an approximate Gaussian distribution according to the central limit theorem as a result of the decorrelation process in the previous stage,and hence the BAVQ can also be used.Moreover,the codebook sharing becomes possible after normalization of those Gaussian distribution data blocks.The test result shows that compared with the traditional MSVQ and BAVQ,this new approach can improve the performance,reduce the calculation burden and take less time.

关 键 词：合成孔径雷达(SAR) 图像压缩 块自适应矢量量化(BAVQ) 多级矢量量化(MSVQ) 码书共享 

分 类 号：TN957.52[电子电信—信号与信息处理]