0
$\begingroup$
  1. Euclidian distance
  2. Cosine similarity
  3. Histogram Intersection
  4. Or any other?

Consider the importance of the method in case of images. Eg, Cosine Similarity matches angle of two vectors...
What impact comparing just the angle or distance has on accuracy of BOW?
Detailed explanation is welcome!!!

$\endgroup$
0
$\begingroup$

Histogram Intersection (Pyramid Match Kernel) specifically is currently accepted as a better tool, but this doesn't generalize really. When you formulate your features or histogram computations differently, things are subject to change.

$\endgroup$
  • $\begingroup$ Can using Histogram Intersection for matching 128-D SIFT descriptor rather than Euclidean distance for object detection give better matches? $\endgroup$ – object recognition Mar 5 '14 at 16:20
  • $\begingroup$ I guess. However, Arandjelovi and Zissermann proposed a nice technique for matching Sift with Hellinger distance and they name it RootSift. I would advise you to read through it : robots.ox.ac.uk/~vgg/publications/2012/Arandjelovic12/… $\endgroup$ – Tolga Birdal Mar 5 '14 at 16:29

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.