基于多DSP的遥感图像实时压缩系统设计

随着遥感技术的发展,对高分辨率的遥感图像实时压缩的需求日益迫切.设计了高性能的图像压缩系统,由8片ADSP-TS201为核心处理器和2片FPGA组成,可提供高达28.8 GFLOPS的峰值浮点运算能力.该平台采用PCIE总线作为外部接口,具有良好的可扩展性和强大的数据交换能力;各处理器采用Link Port互联,形成松耦合结构,可实现高速传输、实时处理和大容量存储三者的平衡,具有良好的可重构性和通用性.最后采用该系统实现JPEG2000压缩,实验结果表明,该压缩系统无损压缩速率可达6.2 Mpixels/s,非常适用于高分辨率、高质量遥感图像压缩领域....

Full description

Saved in:
Bibliographic Details
Published in电子技术应用 Vol. 41; no. 5; pp. 46 - 49
Main Author 贺文静 胡坚 李子扬 孟凡荣 周春城
Format Journal Article
LanguageChinese
Published 中国科学院光电研究院中国科学院定量遥感信息技术重点实验室,北京,100094 2015
Subjects
FPGA
JPEG2000
图像压缩
多DSP
并行处理
multi-DSP
JPEG2000
并行处理
parallel process
FPGA
多DSP
image compression
图像压缩
Online AccessGet full text
ISSN0258-7998
DOI10.16157/j.issn.0258-7998.2015.05.008

Cover

More Information
Summary:随着遥感技术的发展,对高分辨率的遥感图像实时压缩的需求日益迫切.设计了高性能的图像压缩系统,由8片ADSP-TS201为核心处理器和2片FPGA组成,可提供高达28.8 GFLOPS的峰值浮点运算能力.该平台采用PCIE总线作为外部接口,具有良好的可扩展性和强大的数据交换能力;各处理器采用Link Port互联,形成松耦合结构,可实现高速传输、实时处理和大容量存储三者的平衡,具有良好的可重构性和通用性.最后采用该系统实现JPEG2000压缩,实验结果表明,该压缩系统无损压缩速率可达6.2 Mpixels/s,非常适用于高分辨率、高质量遥感图像压缩领域.
Bibliography:The demand for real-time compression of high-resolution remote sensing image is becoming urgent, with the develop- ment of remote sensing technology. A muhi-DSP parallel processing system for real-time image compression is designed. The sys- tem is composed of eight DSPs and two FPGAs, providing 28.8 GFLOPS operations performance. PCI-Express bus is used as exter- nal interface, providing the powerful capability of communication and expansion. To archive the high performance in data transmis- sion,processing and large capacity storage, the loose coupled parallel architecture is used, with processors interconnected by Link Port. An application example of JPEG2000 compression based on this system is given. It is shown that the throughout speed of the system can achieve 6.2 Mpixels/s with lossless compression. Therefore, it is suitable for the compressing application of high-resolu- tion remote sensing image.
11-2305/TN
muhi-DSP ; FPGA ; image compression; JPEG2000; parallel process
He Wenjing, Hu Jian, Li Ziyang,
ISSN:0258-7998
DOI:10.16157/j.issn.0258-7998.2015.05.008