0

I am a freshman of vl_sift. When I ran the codes from the website, I have found a problem which I didn't understand. After executing [f,d] = vl_sift(I) ; I checked the f.

col   11                      12       

15.6534681320190    15.6534681320190
315.868621826172    315.868621826172
1.75902485847473    1.75902485847473
0.729515715938409   -2.98263351880919

Two columns have the same values of X,Y but different orientations. What's the mean? Could anyone help me? Thank you!

Luc M
  • 16,630
  • 26
  • 74
  • 89
Anguslilei
  • 31
  • 1
  • 5

1 Answers1

0

The function that computes the orientation of a given keypoint can indeed compute multiple orientations. In practice vlfeat can return up to 4 orientations for the same keypoint:

For more details, please refer to the section 5. Orientation assignment of the original paper:

Peaks in the orientation histogram correspond to dominant directions of local gradients [...] for locations with multiple peaks of similar magnitude, there will be multiple keypoints created at the same location and scale but different orientations.

deltheil
  • 15,496
  • 2
  • 44
  • 64
  • I wonder that one pixel corresponds to just only one vector of descriptor ,or two?I might not totally understand SIFT Algorithm.Thanks a lot! – Anguslilei Apr 12 '13 at 06:44