I had a real good go at this, but only managed half the effect needed I'm not back into work till tomorrow (GMT+1) but I could show you where I got to. What was the end result to be used for?
The end result is a delaunay edges component with some culling of the edges greater than 1. I'm sure the same theory can be applied to a mesh to get a surface, But I'm not sure how to get the boundary out or if it's possible.
Not exactly the thing you want and only a sketch of a solution, but here's my contribution to the brainstorm, maybe someone can pick it up from there...
Though again - it's not exactly what you wanted - the bounding curve will "jump the gaps" with concave or flat curves only... so you won't get the effect shown on the bottom of your image.
I've been working on a method - it's not working as well as the ones below but maybe the different approach can spark some new ideas...
basically i throw a bounding box around the input curves, draw lines radiating out from the centroid, pick the furthest point along each line & connect 'em. Just a quick sketch.
thanks to everyone for the help and some definitions too.
i'm hoping that this will solve a major flaw in a definition i made to create a planking effect on the curves.
thanks again, and please by all means, keep the ideas flowing. This next definition I make will be the 7th version and everyone of them have been done differently to try and compensate for a few hiccups that I can't figure out why they are failing.