2

I want to normalize the sift descriptors by rotating them so that the horizontal direction is aligned with the dominant gradient orientation of the patch.

I am using vl_feat library. Is there any way in vl_feat to normalize the sift descriptos?

or

what is the effective way of doing this using matlab?

user570593
  • 3,420
  • 12
  • 56
  • 91

1 Answers1

3

I believe the ones in VLfeat are already oriented in the dominant gradient direction.

It shows them rotated if you look here: http://www.vlfeat.org/overview/sift.html 

 [f,d] = vl_sift(I) ;

f is a Nx4 matrix of keypoints. N being the keypoint indexand the other 4 being, x position, y position, scale, and orientation. d is a Nx128 matrix, where N is the keypoint index, and the 128 dimensions belong to the SIFT descriptor.

If all of your images are rotated upright, it is actually beneficial to not use rotational invariance. See this paper which assume a gravity vector: https://dspace.cvut.cz/bitstream/handle/10467/9548/2009-Efficient-representation-of-local-geometry-for-large-scale-object-retrieval.pdf?sequence=1

Erotemic
  • 4,806
  • 4
  • 39
  • 80
  • Hi thanks for the reply. I am not sure about the vl_sift whether it is rotated or not. I am trying to match the feature points from one image with another image. In that case do you think I dont need to rotate the sift features to its dominant orientation? – user570593 Mar 23 '13 at 16:38
  • 2
    You currently have orientation-invariance (because vl_sift does this by default). My second point was: right now, the consensus in the literature is that *if your images are all upright* then it is better to leave the descriptor unrotated. Now, if you want to unrotate your SIFT descriptors you will have to look in the tutorial on the webpage I sent you. VL_sift allows for a Frames argument, taking the keypoints (the fc variable: x,y,scale,orientation) that you want to describe, and computes a sift descrtiptor for each of those. Set orientation to 0 if you want the gravity vector. – Erotemic Mar 23 '13 at 21:28