High-speed parallel wavelet algorithm based on CUDA and its application in three-dimensional surface texture analysis
The State Key Laboratory of Digital Manufacturing Equipment and Technology, School of Mechanical Science and Engineering, Huazhong University of Science and Technology, Wuhan, Hubei, China
International Conference on Electric Information and Control Engineering (ICEICE), 2011
@inproceedings{jianjun2011high,
title={High-speed parallel wavelet algorithm based on CUDA and its application in three-dimensional surface texture analysis},
author={Jianjun, W. and Wenlong, L. and Xiaojun, L. and Xiangqian, J.},
booktitle={Electric Information and Control Engineering (ICEICE), 2011 International Conference on},
pages={2249–2252},
organization={IEEE},
year={2011}
}
A new efficient parallel wavelet algorithm was presented in order to speed up wavelet transform in three-dimensional surface texture analysis. It is based NVIDIA’s CUDA (Compute Unified Device Architecture), a new general purpose parallel programming model and instruction set architecture that leverage computational problems on GPU more efficient than CPU. Compared with CPU, GPU has evolved into a highly parallel, multithread, multicore processor with tremendous computational horsepower and very high memory bandwidth. GPU is well-suited to address data-parallel computation problems rather than flow controlled problems. Wavelet transform can use data-parallel programming model so data elements will be mapped to parallel processing threads to speed up the computations. CUDA wavelet decomposition and reconstruction algorithms were realized based on the analysis above. Experiments show that the parallelization of the fast wavelet decomposition transform for GPU speedup 34x-38x over CPU, reconstruction transform speedup 29x-33x over CPU.
June 25, 2011 by hgpu
