|
Article information
2016 , Volume 21, ¹ 6, p.47-58
Lyubutin P.S., Panin S.V.
The use of parallel computing with the AMD graphics processors for the construction of displacement vector fields
The aim of the study is to develop a parallel algorithm for constructing displacement vector fields based on recursive search approach with the use of parallel computing at GPU and estimation of time costs. A modified algorithm for the displacement estimation DRS (direct recursive search) was implemented beingand based on the approach to three-dimensional recursive search (3DRS). By considering the DRS features and limitations for the hardware parallel computing, the parallel algorithm PDRS for constructing displacement vector fields has been developed. The latter employs recursive search, which was implemented with the use of the OpenCL programming language to run onto GPUs AMD Radeon. Test results showedhave shown that the use of graphics processors can significantly reduce the time of construction of displacement vector fields. Processing time at preset parameters and images with a resolution of 3456 × 5184 was decreasing by ∼ 27 times. When the density of the vector field was increased by three times (spacing between vectors was equal to 16 pixels) as well as the number of vectors in the lines the time required for implementing the parallel algorithm PDRS was 63 times less as compared to the DRS algorithm, realized withwhich was implemented using the high level language C++ for the CPU operation by CPU. This is related to the fact that with increasing the size of the vectors in the line the program uses a larger number of cores in parallel graphics method for constructing vectors in a row as the size of the vectors in the line increases. The vector operations were applied to compute the similarity measure SAD. Their employment incorporation allows reducing the computation time by 1.5-2 times at use offor AMD Radeon 5570 and 7970 graphical cards. In contrast with, the current implementation being realized with, which uses the PDRS algorithm that employs vector operations on the previous generation graphics card Radeon 5570 a previous generation one shows a has shown greater effect of reducing the computation time. The reason is associated with the differences between the architecture of those two video processors.
[full text]
Keywords: recursive displacements search, displacement vector field, parallel algorithm, graphics processing unit
Author(s):
Lyubutin Pavel Stepanovich
PhD.
Position: Junior Research Scientist
Office: Institute of Strength Physics and Materials Science of SB RAS
Address: 634021, Russia, Tomsk
Phone Office: (3822)286-889
E-mail: psl@sibmail.com
Panin Sergey Viktorovich
Dr. , Professor
Position: Head of Laboratory
Office: Institute of Strength Physics and Materials Science of SB RAS, National Research Tomsk Polytechnic University
Address: 634021, Russia, Tomsk
Phone Office: (3822)286-904
E-mail: svp@ispms.tsc.ru
References:
1] Schreier, H., Orteu, J.-J., Sutton, M.A. Image correlation for shape, motion and deformation measurements. Basic concepts, theory and applications. Springer; 2009: 321. DOI:10.1007/978-0-387-78747-3
[2]Raffel, M., Willert, C.E., Wereley, S.T., Kompenhans, J. Particle Image velocimetry. A practical guide. 2nd ed. Springer Berlin Heidelberg; 2007: 448. DOI:10.1007/978-3-540-72308-0
[3] Belyaev, E.A., Tyurlikov, A.M. Motion estimation algorithms for low bit-rate video Compression. Computer Optics. 2008; 32(4):403–412. (In Russ.)
[4] Giachetti, A., Campani, M., Torre, V. The use of optical flow for road navigation. IEEE Transactions on Robotics and Automation. 1998. Vol. 14, No. 1. P. 34–48.
[5] Doyle, D.D., Jennings, A.L., Black, J.T. Optical flow background estimation for real-time pan/tilt camera object tracking // Measurement. 2014. Vol. 48. P. 195–207.
[6] Garrigues, M., Manzanera, A. Real time semi-dense point tracking. Image Analysis and Recognition. 9th International Conference, ICIAR 2012, Aveiro, Portugal, June 25-27, 2012. Proceedings, Part I . Springer Berlin Heidelberg; 2012: 245–252.
[7] Mahmoudi, S.A., Kierzynka, M., Manneback, P., Kurowski, K. Real-time motion tracking using optical flow on multiple GPUs. Bulletin of the Polish Academy of Sciences Technical Sciences. 2014; 62(1):139–150.
[8] Gao, W., Kemao, Q. Parallel computing in experimental mechanics and optical measurement: A review. Optics and Lasers in Engineering. 2011; 50(4):608–617.
[9] Zhang, L., Wang, T., Jiang, Z., Kemao, Q., Liu, Y., Liu, Z., Tang, L., Dong, S. High accuracy digital image correlation powered by GPU-based parallel computing. Optics and Lasers in Engineering. 2015; (69):7–12.
[10] Shao, X., Dai, X., He, X. Noise robustness and parallel computation of the inverse compositional Gauss–Newton algorithm in digital image correlation. Optics and Lasers in Engineering. 2015; (71):9–19.
[11] Barranco, F., Dıaz, J., Pino, B., Ros, E. A multi-resolution approach for massively-parallel hardware-friendly optical flow estimation. Journal of Visual Communication and Image Representation. 2012; 23(8):1272–1283.
[12] Garcia-Dopico, A., Pedraza, J.L., Nieto, M., P´erez, A., Rodr´ıguez, S., Navas, J. Parallelization of the optical flow computation in sequences from moving cameras. Eurasip Journal on Image and Video Processing. 2014. http://jivp.eurasipjournals.springeropen.com/articles/10.1186/1687-5281-2014-18 DOI: 10.1186/1687-5281-2014-18.
[13] Shiralkar, M.P., Schalkoff, R.J. A self-organization based optical flow estimator with GPU Implementation. Machine Vision and Applications. 2012; 23(6):1229–1242.
[14] Sundaram, N., Brox, T., Keutzer, K. Dense point trajectories by GPU-accelerated large displacement optical flow. Computer Vision — ECCV 2010. Proceedings of the 11th European Conference on Computer Vision: Part 1. Heraklion, Crete, Greece, September 5–11, 2010. Lecture Notes in Computer Science. 2010; (6311):438–451.
[15] Plyer, A., Le Besnerais, G., Champagnat, F. Massively parallel Lucas Kanade optical flow for real-time video processing applications. Journal of Real-Time Image Processing. 2016; 11(4):713–730.
[16] de Haan, G., Biezen, P.W.A.C., Huijgen, H., Ojo, O.A. True-motion estimation with 3-D recursive search block matching. IEEE Transactions on Circuits and Systems for Video Technology. 1993; 3(5):368–379.
[17] Zhao, M., van der Heijden, H. 3D recursive search block matching on graphics processing Unit. Consumer Electronics, 2008. ICCE 2008. Digest of Technical Papers. International Conference on. 9-13 Jan. 2008. http://ieeexplore.ieee.org/document/4587939/?arnumber=4587939 DOI: 10.1109/ICCE.2008.4587939.
[18] Gray, K. Microsoft DirectX 9 programmable graphics pipeline. Microsoft Press; 2003: 496.
[19] St-Laurent, S., Engel, W. The complete effect and HLSL guide. London: Paradoxal Press; 2005: 324.
[20] Panin, S.V., Titkov, V.V, Lyubutin, P.S. Computationally effective three dimensional recursive algorithm for calculation of the translation vectors in the optical method for assessment of deformations. Computational Technologies. 2013; 18(5):91–101. (In Russ.)
[21] Aaftab Munshi, Benedict Gaster, Timothy Mattson, James Fung, Dan Ginsburg OpenCL programming guide. Addison-Wesley Professional; 2011: 648.
Bibliography link:
Lyubutin P.S., Panin S.V. The use of parallel computing with the AMD graphics processors for the construction of displacement vector fields // Computational technologies. 2016. V. 21. ¹ 6. P. 47-58
|
