抄録
K-054
Task level SURF keypoint extraction parallelization in AR application
◎Hoang Anh Dang・Pao Sriprasertsuk・亀山　渉（早大）
In an effort of developing a flexible AR system that is unconstrained by environment, we have implemented parallel SURF key point extraction in which smaller parts of the image are processed simultaneously. Experiments show that the new system is more suitable for new generation of hardware and result lower processing time. In this paper, we discuss various aspects of this implementation: the consistency of extracted key point compared to the traditional SURF implementation, the optimum safe-margin of image slices, the relationship between performance and the number of image slices.