3

I know how to compare 2 images with SIFT (in matlab) but what if I have to find the similarity in between multiple images. Let for instance, I have 30 Images. How can I use SIFT to say that 10 of them have 30% match and 5 of them have 95% match? I can use clustering but before doing that I need some kind of value on which I can apply clustering technique. I have been bothered by this question for quiet a few hours. So friends if you know please do help me.

thanks

Amro
  • 123,847
  • 25
  • 243
  • 454
Sajan
  • 121
  • 10
  • Consider the bag of words model. See my reply here: [http://stackoverflow.com/questions/4357352/euclidean-distance-in-sift/8549874#8549874][1] [1]: http://stackoverflow.com/questions/4357352/euclidean-distance-in-sift/8549874#8549874 – lightalchemist Dec 18 '11 at 04:50

1 Answers1

1

Consider SIFT-ing by comparison to the average image, then cluster by similar amounts/locations of difference across the image. Then, you can compare the image clusters to the cluster-averages.

bythenumbers
  • 155
  • 1
  • 3
  • 10