[Insight-users] Re: Re: 2D boundary tracking

[Zachary Pincus]

> One more detail. From the first few web pages I found describing
> the algorithm, it seems the output contour is not a path
> (just a collection of possibly non-adjacent little segments).
> From your documentation I gather that the output is a directed
> path, with multiple components if need be, and for that you have
> to go to great extents, and use the two hash tables, to keep
> track what piece goes where. Is that correct? That's pretty clever.

Correct. The output is a set of directed paths, each comprising a set  
of correctly-linked segments. (It's easy to figure out the correct  
direction of the segments by keeping the low side of the image  
function always on the left or the right; then all that's needed is  
to properly link up segments with the same beginning or ending  
points.) Some of the paths might be open if they intersect the image  
edge; otherwise they will be closed.

Zach