摘要
讨论了多维 DCT和多维 DWT的并行行列算法和并行多项式变换算法 ,并用 Log P模型对算法进行了分析。在仔细分析一维和二维离散小波变换与小波包变换计算结构的基础上 ,提出了它们的并行算法。算法只需相对较少的通信时间 ;适合大规模并行机 ( MPP)和工作站或微机机群系统 ;方法也适合信号处理中的各种塔式分解过程。用 Fortran语言和 PVM编制了算法的程序。在机群系统和大规模并行机上的实验表明 ,算法是正确的且具有较高的加速比。
The parallel row column method and the parallel polynomial transform algorithm are discussed. The algorithms are analyzed with the LogP model. After detailed analysis of the structure of one and two dimensional discrete wavelet transform, the paper proposes parallel algorithms for them. The algorithms need relatively small amount of communication time and are suitable for MPP or workstation clusters. Programs are made for them. Experiments on MPPs show that the algorithms are correct and high speedups are achieved. The methods can also be used for any kind of pyramid a Algorithm in signal processing.
出处
《国防科技大学学报》
EI
CAS
CSCD
2000年第2期41-45,共5页
Journal of National University of Defense Technology
基金
国家自然科学基金资助项目! ( 1960 10 12 )
关键词
并行算法
离散变换
小波
塔式算法
信号处理
Parallel algorithm
Discrete Wavelet Transform
Pyramid Algorithm
Signal Processing
