Correlation pattern recognition

Hi,

(hope this is the right forum to ask)

suppose I have a 2D scatter plot plotting some correlation between two variables.
Suppose that due to some underlying processes the plot shows a structure having,
say, two set of points/clusters (maybe overlapping) with different correlation coefficient.

Is there some multivariate tool in ROOT/RooFit/etc with which one can recognize such a pattern
in a 2D histo/image and sort the points (perhaps in some probabilistic way) automatically according to which correlation they correspond to?

Or is there some method name you could suggest what one may try to search for in the literature?

Thanks for any ideas,
Balint

Answering to my own question :slight_smile:

I realized TMVA is my friend, although I am still interested in whether general image pattern
recognition algorithms (using things like Fourier-, Laplace- or Hough-transformation, etc.) are implemented somewhere in the ROOT class-space.

Thanks,
Balint