改进的零树编码算法及其DSP实现

Modified Zerotree Coding Algorithm and Its DSP Implementation

基于双正交叠式变换(LBT)的卫星遥感图像数据压缩,需要一种低内存和低复杂度的编码方法。以SP IHT(基于分层树的集合分割排序的编码算法)为基础,通过研究LBT系数的结构,提出了一种改进的零树编码算法。该算法采用了零树结构的快速检验和线性索引技术。本文对该算法在TM S320C 6000系列DSP中的实现进行了研究,通过任务合并、优化选项、软件流水、循环展开和多功能指令等方法提高处理速度。该算法已在某侦查卫星数据压缩样机中实现。实验结果表明,该算法内存需求低,压缩质量接近SP IHT,并且计算复杂度大大降低。

The satellite remote sensing image compression based on lapped biorthogonal transform （LBT） requires a low-complexity and low-memory coder. On the basis of set partitioning in hierarchical trees（SPIHT）, the structure of LBT coefficients is analyzed and a modified zerotree coding algorithm uses techniques of the zerotree structure speediness inspection and the linear index. Furthermore, the implementation in TMS320C6000 DSP of the algorithm is studied. The processing speed is improved by task combining,compiler options,software pipelining,loop unrolling, and complex instruction. The algorithm is used in a prototype of data compression equipment in a certain spy satellite. Image coding experiments show that the memory requirement of the algorithm is low, and its coding performance is competitive with that of SPIHT while its complexity considerably decreases.