Abstract
SIFT, short for Scale Invariant Feature Transform, is regarded as one of the most robust feature detection algorithms. The Gaussian and DoG Pyramid Construction part, functioning as computation basis and searching spaces for other parts, proves fatal to the system. In this paper, we present an FPGA-implementable hardware accelerator for this part. Stratified Gaussian Convolution scheme and 7-Round Parallel Computation scheme are introduced to reduce the hardware cost and improve process speed, meanwhile keeping high accuracy. In our experiment, our proposal successfully realizes a system with max clock frequency of 95.0 MHz, and on-system process speed of up to 21 fps for VGA format images. Hardware cost of Slice LUTs is reduced by 12.1% compared with traditional work. Accuracy is kept as high as 98.27% against original software solution. Our proposed structure proves to be suitable for real-time SIFT systems.
This is a preview of subscription content, log in via an institution.
Buying options
Tax calculation will be finalised at checkout
Purchases are for personal use only
Learn about institutional subscriptionsPreview
Unable to display preview. Download preview PDF.
References
Lowe, D.G.: Object recognition from local scale-invariant features. In: ICCV 1999 Proceedings, Corgu, Greece, pp. 1150–1157 (1999)
Lowe, D.G.: Distinctive image features from scale-invariant keypoints. IJCV 60(2), 91–110 (2004)
Dermitzakis, K., McKenzie, E.: A gpu implementation of the sift algorithm: An msc project proposal (2007), www.inf.edu.ac.uk/events/jamboree/2007/Posters/k-dermizakis.dpf
Sinha, S.N., Frahm, J.-M., Pollefeys, M., Genc, Y.: Gpu-based video feature tracking and matching. Technical Report TR, Department of Computer Science, UNC Chapel Hill 06(012) (2006)
Heymann, S., Muller, K., Smolic, A., Frohlich, B., Wiegand, T.: Sift implementation and optimization for general-purpose gpu. In: WSCG 2007 Proceedings, University of West Bohemia, Plzen, Campus Bory, p. G03 (2007)
Qiu, J., Huang, T., Ikenaga, T.: Hardware accelerator for feature point detection part of sift algorithm and corresponding hardware-friendly modification. In: SASIMI 2009 Proceedings, Okinawa, Japan, pp. 213–218 (2009)
Qiu, J., Huang, T., Ikenaga, T.: 1d-based 2d gaussian convolution unit based hardware accelerator for gaussian dog pyramid construction in sift. In: IEICE 2009 Proceedings, Matsuyama, Japan, pp. 1150–1157 (2009)
Qiu, J., Huang, T., Huang, Y., Ikenaga, T.: A hardware accelerator with variable pixel representation and skip mode prediction. In: MVA 2009 Proceedings, Tokyo, Japan, pp. 1150–1157 (2009)
Bonato, V., Marques, E., Constantinides, G.A.: A parallel hardware architecture for scale and rotation invariant feature detection. IEEE trans. on CSVT 18(12), 1703–1712 (2008)
Se, S., Ng, H.K., Jasiobedzki, P., Moyung, T.J.: Vision based modeling and localization for planetary exploration rovers. In: ICA 2004, p. 11 (2004)
Rob hess - school of eecs @ oregon state university, http://web.engr.oregonstate.edu/hess/
Kawasaki, H., Furukawa, R.: Shape reconstruction from cast shadows using coplanarities and metric constraints. In: Yagi, Y., Kang, S.B., Kweon, I.S., Zha, H. (eds.) ACCV 2007, Part II. LNCS, vol. 4844, pp. 847–857. Springer, Heidelberg (2007)
Author information
Authors and Affiliations
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2010 Springer-Verlag Berlin Heidelberg
About this paper
Cite this paper
Qiu, J., Huang, T., Ikenaga, T. (2010). A 7-Round Parallel Hardware-Saving Accelerator for Gaussian and DoG Pyramid Construction Part of SIFT. In: Zha, H., Taniguchi, Ri., Maybank, S. (eds) Computer Vision – ACCV 2009. ACCV 2009. Lecture Notes in Computer Science, vol 5996. Springer, Berlin, Heidelberg. https://doi.org/10.1007/978-3-642-12297-2_8
Download citation
DOI: https://doi.org/10.1007/978-3-642-12297-2_8
Publisher Name: Springer, Berlin, Heidelberg
Print ISBN: 978-3-642-12296-5
Online ISBN: 978-3-642-12297-2
eBook Packages: Computer ScienceComputer Science (R0)