Skip to Main Content
This paper describes design and implementation of a parallel algorithm that detects scale-invariant keypoints from digital images. Keypoint detection is one of the most important operations in image processing since efficient image matching is possible with it. Image matching is used in many intelligent image processing service, including object/scene recognition, stereo correspondence, motion tracking, and panorama imaging. In this paper, we design a new parallel algorithm of keypoint detection that is suitable for Cell Broadband Engine architecture by converting each subprocess of keypoint detection algorithm into parallel version. The experimental results show that performance of our algorithm increases in proportion to the number of processors utilized, that is, it achieves linear scalability.