this is a demo of the sidcell algorithm [1] to compute image correspondences highly efficient.
frame to frame correspondences are merged to an id-tagged list of longterm correspondences.
gpu computation time is lower than 3ms on a tesla c1060, image size 752x480.
[1] Efficient Keypoint Matching for Robot Vision using GPUs, ICCV 2009, 5th IEEE ECVW, Kyoto
Link to this comment:
All Comments (0)