DBScan, Homogeneity and Completeness algorithm

[NoCforMe]

Judging from those pictures, you've cooked up a pretty good "sharpen" filter there, yes?

[guga]

Hi NoCforMe. Yes, the algorithm can be used as a filter to sharpen images. It wasn't designed for that, but in the end I saw that this was another possible use.

Testing for edition. Working now, Stoo :)

[guga]

Finally succeeded to make this works

I created 2 specific functions to properly identify the clusters. One biased on the "normal" identification method as described here: https://www.youtube.com/watch?v=-p354tQsKrs

And other forcing the point nearby a cluster to be a border even if it don´t have the minimum distance of Epsilon. On such cases, if i decide to use this method too (after testing on real images), i´ll see if i can set a newer flag called NEAR_BORDER meaning that we have a point near a cluster that can be considered border despite it´s values is not enough to be considered as such.

Both methods uses the technique described at: https://biswajit.netlify.app/pages/dip-burdwanuniversity

I´ll make further tests before the next release.

This algorithm can be usefull to create a more accurate denoise filter, since it identifies precisely what is noise and also store the necessary amount of data needed for that specific pixel be considered at least a border and later be used to form new clusters if necessary.

Now i need also to find a way to assign each border to it´s own cluster (or clusters) and create a type of subgroup of clusters (for cases where clusters can be merged to form a larger one)

[guga]

Ok, guys. I suceeded to assign a unique ID for each Core and Border points and also succeeded to identify each Noise point as well. Now i´m trying to figure it out a way to assign a unique Id for each cluster and associate it to the corresponding core and border points that forms it.

Any ideas ?

[guga]

Say that i have an image (width = 6, height = 5) that contains these classifications:

Noise - Core - Core - Noise - Core - Core
Border - Border - Core - Border - Border - Core
Noise - Border - Core - Border - Border - Border
Noise - Border - Border - Border - Border - Core
Noise - Noise - Noise - Noise - Core - Core

How do i determine the clusters and how many are there ?