A Hardware Architecture for the Affine-Invariant Extension of SIFT

Affine-invariant extension of scale-invariant feature transform (ASIFT) algorithm requires a large amount of computation and memory access, and consequently, is hard to process in real time. In order to increase the operation speed of ASIFT algorithm, this paper proposes a new hardware architecture for the ASIFT algorithm. In order to reduce the memory access time, the affine transform is modified to allow external memory access in the raster-scan order with a little accuracy drop. In addition, image filtering with skewed kernel is proposed in order to reduce the memory space for image storage. Additional complexity reduction is attempted to reduce the number of simulated viewpoints. As a result, throughput of the affine transform module is increased to 325% and the proposed hardware processes a video graphics array-sized ( $640\times480$ ) video at 20 fps.

[1]  Hyuk-Jae Lee,et al.  A Novel Hardware Architecture With Reduced Internal Memory for Real-Time Extraction of SIFT in an HD Video , 2016, IEEE Transactions on Circuits and Systems for Video Technology.

[2]  Tian-Sheuan Chang,et al.  Low memory cost bilateral filtering using stripe-based sliding integral histogram , 2010, Proceedings of 2010 IEEE International Symposium on Circuits and Systems.

[3]  Luigi Di Benedetto,et al.  Frame buffer-less stream processor for accurate real-time interest point detection , 2016, Integr..

[4]  Jean-Michel Morel,et al.  ASIFT: A New Framework for Fully Affine Invariant Image Comparison , 2009, SIAM J. Imaging Sci..

[5]  Cordelia Schmid,et al.  A Comparison of Affine Region Detectors , 2005, International Journal of Computer Vision.

[6]  Hyuk-Jae Lee,et al.  A novel hardware design for SIFT generation with reduced memory requirement , 2013 .

[7]  Thomas Wiegand,et al.  SIFT Implementation and Optimization for General-Purpose GPU , 2007 .

[8]  Matthijs C. Dorst.  Distinctive Image Features from Scale-Invariant Keypoints , 2011 .

[9]  Normand Teasdale,et al.  Real-time eye blink detection with GPU-based SIFT tracking , 2007, Fourth Canadian Conference on Computer and Robot Vision (CRV '07).

[10]  George A. Constantinides,et al.  A Parallel Hardware Architecture for Scale and Rotation Invariant Feature Detection , 2008, IEEE Transactions on Circuits and Systems for Video Technology.

[11]  Robin Hess.  An Open-Source SIFT Library , 2010 .

[12]  Yung-Chang Chen,et al.  High-Performance SIFT Hardware Accelerator for Real-Time Image Feature Extraction , 2012, IEEE Transactions on Circuits and Systems for Video Technology.

[13]  Jos B. T. M. Roerdink,et al.  GPU-ASIFT: A fast fully affine-invariant feature extraction algorithm , 2013, 2013 International Conference on High Performance Computing & Simulation (HPCS).