Title :
GPU Accelerated Wavelet Transform Profilometry
Author :
Zhou, Bo ; Zhao, Xiaomin ; Liu, Chunyuan
Author_Institution :
Coll. of Comput., Heilongjiang Univ. of Sci. & Technol., Haerbin, China
Abstract :
Huge workload and time-consuming of the phase computation based on the Wavelet Transform Profilometry (WTP) so that not meet real-time three-dimensional (3D) measurement needs. Fortunately the pixels which in situ need to be processed and the already processed pixels are independent, so the phase of each pixel can be calculated separately. It is possible to make the phase computation implement on Graphics Processing Unit (GPU) with the multi-threading parallel processing, and to solve the problem of the low efficiency on CPU. Experiments show that it obtains the speedup of one to two orders of magnitude to CPU by GPU acceleration with the same accuracy, which lays the solid foundation for WTP applying to the real-time 3D measurement.
Keywords :
graphics processing units; multi-threading; parallel processing; wavelet transforms; GPU; WTP; graphics processing unit; multi-threading parallel processing; phase computation; real-time three-dimensional measurement; wavelet transform profilometry; Acceleration; Continuous wavelet transforms; Graphics processing unit; Instruction sets; Real time systems; Compute Unified Device Architecture(CUDA); GPU acceleration; Wavelet Transform Profilometry; computing the phase in parallel;
Conference_Titel :
Computational Sciences and Optimization (CSO), 2012 Fifth International Joint Conference on
Conference_Location :
Harbin
Print_ISBN :
978-1-4673-1365-0
DOI :
10.1109/CSO.2012.24