Publication Details

GPU Optimization of Convolution for Large 3-D Real Images

KARAS Pavel, SVOBODA David and ZEMČÍK Pavel. GPU Optimization of Convolution for Large 3-D Real Images. In: Proceedings of ACVIS 2012. Heidelberg: Springer Verlag, 2012, pp. 59-71. ISBN 978-3-642-33139-8.
Czech title
Optimalizace konvoluce velkých reálných 3D obrazů na GPU
Type
conference paper
Language
english
Authors
Karas Pavel, RNDr. (FI MUNI)
Svoboda David, doc. RNDr., Ph.D. (FI MUNI)
Zemčík Pavel, prof. Dr. Ing. (DCGM FIT BUT)
Keywords

gpu, convolution, 3-D, image processing

Abstract

In this paper, we propose a method for computing convolution of large 3-D images with respect to real signals. The convolution is performed in a frequency domain using a convolution theorem. Due to properties of real signals, the algorithm can be optimized so that both time and the memory consumption are halved when compared to complex signals of the same size. Convolution is decomposed in a frequency domain using the decimation in frequency (DIF) algorithm. The algorithm is accelerated on a graphics hardware by means of the CUDA parallel computing model, achieving up to 10x speedup with a single GPU over an optimized implementation on a quad-core CPU.

Published
2012
Pages
59-71
Proceedings
Proceedings of ACVIS 2012
Conference
Advanced Concepts for Intelligent Vision Systems (ACIVS) 2012, Faculty of Information Technology, Brno University of Technology, Brno, CZ
ISBN
978-3-642-33139-8
Publisher
Springer Verlag
Place
Heidelberg, DE
DOI
10.1007/978-3-642-33140-4_6
BibTeX 
@INPROCEEDINGS{FITPUB10243,
   author = "Pavel Karas and David Svoboda and Pavel Zem\v{c}\'{i}k",
   title = "GPU Optimization of Convolution for Large 3-D Real Images",
   pages = "59--71",
   booktitle = "Proceedings of ACVIS 2012",
   year = 2012,
   location = "Heidelberg, DE",
   publisher = "Springer Verlag",
   ISBN = "978-3-642-33139-8",
   doi = "10.1007/978-3-642-33140-4\_6",
   language = "english",
   url = "https://www.fit.vut.cz/research/publication/10243"
}
Back to top