Color Image Segmentation with Dbscan

Hi,

I am about to finde clusters in a image with the DBScan. To do so I have tried to adapt the workflow of christian.dietz from the entry “Applying cluster nodes on images”. Basically, I want to do what’s described here: https://pdfs.semanticscholar.org/b6d5/5f86aca5e6571d5742fdd360f401634b9401.pdf
As I understand it, the main differences are that it is done by using HSV/HSB and that i want to use DBScan instead of k-means. Here is a example what it should look like:


I just replaced the k-means node with numeric distances and DBSCAN but it not working. Is that even the right approach?

Thanks for advice.

DBSCAN is recommended to look for outliers. Is it your case?

Yes I need to do it with DBscan