Change distance function for kmeans clustering

jstr
6 ●2

Hi there,

I have a question concerning the kmeans method for clustering data-points. By default this method uses the L2 norm (euclidean distance) to cluster the provided data. Is there any possibility for using another distance metric?

add a comment

1 answer

Sort by » oldest newest most voted

answered Jan 10 '13

Alexander Smorkalov

5862 ●2 ●22 ●58 http://alex-alchemist....

No, you cannot use another method for kmeans without OpenCV library modification. kmeans function calls normL2Sqr_ function from stat.cpp for distance calculation. You can change this call on your function with the same signature.

PS: Willowgarage pages with OpenCV documentation is outdated. Use official OpenCV documentation at http://docs.opencv.org/

Comments

It would be good if the outdated documentation at least contains a link to the actual one. It is really easy to find old versions of the docs. For instance googling for cvtColor shows willowgarage docs at 2nd and 3rd place.