Title :
SIFT Implementation Based on Parallel Computation
Author :
Feng Liang ; Xu Tingfa ; Huang Qian ; Wang Xiaoyan ; Wang Ping
Author_Institution :
Key Lab. of Photoelectronic Imaging Technol. & Syst., Beijing Inst. of Technol., Beijing, China
Abstract :
Image matching is the basis of object recognition, tracking and three-dimensional reconstruction. The SIFT algorithm based on scale space has been widely used for image matching due to its strong robustness. However, the performance of the original flow is difficult to satisfy the real-time requirements. Due to the great advantage of parallel computing, this paper modified the algorithm flow of SIFT algorithm. The experimental results show that the modified algorithm flow makes the computation faster but can find more effective feature points. Also, the experiments further prove the invariance to change of image scales and rotation angles still exists apparently.
Keywords :
image matching; image reconstruction; object recognition; real-time systems; SIFT algorithm; feature points; image matching; object recognition; object tracking; parallel computing; real-time requirement; scale invariant feature transform; three-dimensional reconstruction; Computer vision; Feature extraction; Graphics processing unit; Image edge detection; Image matching; Object recognition; Real time systems;
Conference_Titel :
Wireless Communications Networking and Mobile Computing (WiCOM), 2010 6th International Conference on
Conference_Location :
Chengdu
Print_ISBN :
978-1-4244-3708-5
Electronic_ISBN :
978-1-4244-3709-2
DOI :
10.1109/WICOM.2010.5600662