I need some advice regarding the automatic filtering out of noisy / high-variance parts of a dataset. Here is a dropbox link to an illustration of the problem: Link
Basically the dataset (on the left) is noisy / high-variance towards the rim and I want to design an automatic routine to cut away that region. While this figure illustrates a typical situation, the scaling of both data and variance, as well as line between high/low variance may vary between datasets so I am a bit stuck on how to automate the cutaway.
I appreciate any advice you can give me.