SAR原始数据压缩技术研究

Study on SAR Raw Data Compression Techniques

研究两种合成孔径雷达(SAR)原始数据压缩算法,它们是块自适应树型矢量量化算法和块自适应预测编码算法。前者是在使用穷尽型搜索技术的块自适应矢量量化算法的基础上,通过使用树型搜索算法来提高算法的运行效率;后者是通过预测编码来消除SAR原始数据之间的相关性从而提高压缩性能。结合机载SAR实测原始数据,对讨论的各种算法分别进行压缩和解压缩,并进行SAR成像处理。通过比较和分析各种算法的性能及图像域参数,表明块自适应树型矢量量化算法和块自适应预测编码算法能提高SAR原始数据的压缩性能,比较适合实际工程应用。

Two raw data compression algorithms are applied to synthetic aperture radar (SAR) raw data. The block adaptive tree-structure vector quantization (BATSVQ) and the block adaptive predictive quantization (BAPQ) are analyzed. To compare the computational load of BATSVQ with full search block adaptive vector quantization (BAVQ), the LBG algorithm is used to generate a code book for full search. The BATSVQ outperforms the BAVQ at the same rate. Because a linear predictor with few taps can capture most of raw signal correlation, the performance of the BAPQ is superior to that of the block adaptive quantization (BAQ). By using the airborne SAR raw data, the performances achieved in terms of bit reduction and certain quality parameters in the image domain have been evaluated. The BATSVQ and BAPQ algorithms seems well-suited to raw SAR data compression in terms of computational complexity and quality of encoded image at low rates.