摘要
卫星数据压缩要求实时压缩、低比特率和良好的图像质量,系统压缩算法选择基于定点叠式变换的低复杂度图像压缩方法,给出了定点叠式变换的实现流程和一种改进的零数编码的思路;在综合考虑成本、功耗、数据接口和处理能力的基础上,提出了多DSP+FPGA的硬件方案,主要由DSP完成压缩运算,型号选取TMS320C6416,对压缩算法的DSP实现中的一些关键问题,如数据溢出、代码优化、并行计算等技术进行了研究。以一组遥感灰度图像为样本,分4:1和8:1两种压缩比对数据压缩模块进行了测试;在CCD相机时钟频率为50MHz、帧频为1.3K、每帧4096字节的输出条件下,每个DSP可以处理两路图像数据,压缩图像质量良好。
Satellite data compression requires real time processing, low bit rate and high fidelity of reconstruct image. The compression arithmetic is a low complexity compression method based on fixed point lapped biorthogonal transform. The flow chart of fixed point lapped biorthogonal transform and the thing of a modified zero tree coding algorithm are gived. Based on the analysis to cost. energy consume, interface and performance, the project of hardware based on multi-DSP and FPGA is put forward. The computation is mainly done by DSP whose type is TMS320C6416. Also, some key techniques on transplant of compression arithmetic from PC to DSP, such as data overflow, optimization methods for the code and parallel computing are studied. The data compression module is tested at the ratio of 4：1 and 8：1 based on a group of remote sensing images. Every DSP can deal with picture datas of two channel when the clock frequency of CCD camera is 50M, the frame frequency is 1. 3K and every frame has 4096 bytes, the reconstruct image quanlity is good.
出处
《计算机测量与控制》
CSCD
2007年第8期1090-1092,共3页
Computer Measurement &Control
基金
国防科技工业基础科研项目(H142005D001)。
关键词
数据压缩
定点叠式变换
多DSP+FPGA
并行计算
data compression
fixed point lapped biorthogonal transform
multi-DSP and FPGA
parallel computing
