基于OpenCL的JPEG压缩算法并行化设计与实现

随着图像数据的大量增加，传统单处理器或多处理器结构的计算设备已无法满足实时性数据处理要求。异构并行计算技术因其高效的计算效率和并行的实时性数据处理能力，正得到广泛关注和应用。利用GPU在图形图像处理方面并行性的优势，提出了基于OpenCL的JPEG压缩算法并行化设计方法。将JPEG算法功能分解为多个内核程序，内核之间通过事件信息传递进行顺序控制，并在GPU4-CPU的异构平台上完成了并行算法的仿真验证。实验结果表明，与CPU串行处理方式相比，本文提出的并行化算法在保持相同图像质量情况下有效提高了算法的执行效率，大幅降低了算法的执行时间，并且随着图形尺寸的增加，算法效率获得明显的提升。...

Full description

Saved in:
Bibliographic Details
Published in计算机工程与科学 Vol. 39; no. 5; pp. 855 - 860
Main Author 张敏华 张剑贤 裘雪红 周端
Format Journal Article
LanguageChinese
Published 西安电子科技大学计算机学院,陕西西安,710071 2017
Subjects
JPEG
OpenCL
并行化
异构计算
heterogeneous computing
JPEG
并行化
OpenCL
parallel processing
异构计算
Online AccessGet full text
ISSN1007-130X
DOI10.3969/j.issn.1007-130X.2017.05.006

Cover

More Information
Summary:随着图像数据的大量增加，传统单处理器或多处理器结构的计算设备已无法满足实时性数据处理要求。异构并行计算技术因其高效的计算效率和并行的实时性数据处理能力，正得到广泛关注和应用。利用GPU在图形图像处理方面并行性的优势，提出了基于OpenCL的JPEG压缩算法并行化设计方法。将JPEG算法功能分解为多个内核程序，内核之间通过事件信息传递进行顺序控制，并在GPU4-CPU的异构平台上完成了并行算法的仿真验证。实验结果表明，与CPU串行处理方式相比，本文提出的并行化算法在保持相同图像质量情况下有效提高了算法的执行效率，大幅降低了算法的执行时间，并且随着图形尺寸的增加，算法效率获得明显的提升。
Bibliography:43-1258/TP
heterogeneous computing ; OpenCL; parallel processing ; JPEG
As the scale of information data increases enormously, traditional single-processor or mul- tiprocessor structure based computing devices are unable to meet the requirements of real-time data pro- cessing. The heterogeneous parallel computing technology attracts much attention and is widely applied for its effective computation efficiency and parallel real-time processing capability. We propose a parallel design of the JPEG compression algorithm based on OpenCL by using the advantages of the GPU in im- age processing. The JPEG algorithm is divided into multiple kernel programs, and the kernels are se- quentially controlled by the event information transfer. The parallel algorithm is simulated and verified on the GPU＋CPU platform. Experimental results show that under the same image quality condition, the parallel algorithm can improve algorithm implementation efficiency and reduce time substantially. And as the graph size increases, the effi
ISSN:1007-130X
DOI:10.3969/j.issn.1007-130X.2017.05.006