Interesting - so this is a sort of boolean for each constituent surface? Not really what I had in mind but I could learn something from that. Maybe one could tie the centroids together into a closed curve, project onto the surface and then extrude in Z. Also messy, and I'm not sure how you would take the outermost delimited surfaces. thanks for taking a stab at it...
so again, there's no direct translation of the gradient and its thresholds in the curvature analysis and real numbers that could generate geometry - is that too ambitious? just as an aside, what renderer are you using...im new to rhino, trying to see what ppl are using...
