Currently, if the feature is, e.g., donut-shaped (or arc-shaped), you can end up with a feature point that is outside of the thresholded region. This is often not ideal; we have to have a whole FAQ page on segmentation about this. It would be good to figure out a technique to pick the "least worst" position within the thresholded region.
Probably the best way to do that would be to have a list of candidate points (all points within the thresholded region) and then pick the closest to what we want. This should be an optional check.
Currently, if the feature is, e.g., donut-shaped (or arc-shaped), you can end up with a feature point that is outside of the thresholded region. This is often not ideal; we have to have a whole FAQ page on segmentation about this. It would be good to figure out a technique to pick the "least worst" position within the thresholded region.
Probably the best way to do that would be to have a list of candidate points (all points within the thresholded region) and then pick the closest to what we want. This should be an optional check.