# Clustering DBSCAN's parameter Epsilon: How is eps related to scale of data being clustered?

How is scale of eps related to data to be clustered in DBSCAN?

e.g. in image of 1024x1024, we have points as:

[[672, 525],
[664, 703],
[632, 717],
[606, 740],
[635, 751]]


If I have set of (x,y) positions to be clustered how should I choose eps? suppose these (x,y) are pixel locations in image and I want to form clusters of (x,y) points, which are 30 pixel or near. i.e. any point can grow its cluster if it has any point within 30 pixel from itself. How should I choose eps? should it be 30? or it depends on my dataset?

I have seen people using eps from 10s/100s/1000s and most popular sub-zero values like 0.2 or something like that.

How should be choose eps if we know the exact units of nearness in out clusters?