Are there similar loop subdivision algorithms for other types of polygon meshes? If I were to guess, would imagine for quadrilaterals subdivision would look something like splitting the shape into four pieces by drawing lines between the midpoints of opposite edges. Furthermore, are there other types of subdivision algorithms for triangle meshes?
Michael-Equi
After going through this slide I feel I still do not have a good intuition for what this subdivision scheme is doing. Perhaps animating the subdivision of the shape could be helpful. Also its wasn't clear to me what it meant by old vertices and new vertices
Edge7481
I'm having a very hard time wrapping my head around what these weights mean. I can tell that they influence the position of the new vertex, (the white one??? I'm not too sure), but what's the point of assigning weights to old vertices? is it keeping, discarding, or moving them?
danielhsu021202
I found a nice animation! Although they use a different averaging equation; I think everything else checks out besides that.
Here's the chapter for loop subdivision, although I suggest watching the whole video, since he doesn't really animate the loop subdivision part.
danielhsu021202
@theflyingpie For quads, we use Catmull-Clark subdivision.
weszhuang
With the way the old vertices are updated with local averaging, does the local averaging need to occur after all points on the mesh are loop subdivided before averaging is performed to prevent weird ordering dependencies on resultant geometry?
Mehvix
@weszhuang you need to use the original neighbor positions when updating old vertices, otherwise you'd end up over-averaging/smoothing. You subdivide first since we want to subdivide on the original, non-smoothed points.
0-0-00-0
I've been wondering why the update for an old vertex uses only its surrounding old vertices, as opposed to using the new vertices (which seems to make more sense to me because the update happens as a consequence of the split). But now I've realized the new vertices are completely determined by the old vertices, and that it would have the same effect to use old ones as using the new ones. The former is also more practical for computation.
Are there similar loop subdivision algorithms for other types of polygon meshes? If I were to guess, would imagine for quadrilaterals subdivision would look something like splitting the shape into four pieces by drawing lines between the midpoints of opposite edges. Furthermore, are there other types of subdivision algorithms for triangle meshes?
After going through this slide I feel I still do not have a good intuition for what this subdivision scheme is doing. Perhaps animating the subdivision of the shape could be helpful. Also its wasn't clear to me what it meant by old vertices and new vertices
I'm having a very hard time wrapping my head around what these weights mean. I can tell that they influence the position of the new vertex, (the white one??? I'm not too sure), but what's the point of assigning weights to old vertices? is it keeping, discarding, or moving them?
I found a nice animation! Although they use a different averaging equation; I think everything else checks out besides that.
https://youtu.be/kC8jbGSiuIQ?t=252&si=7SI5bVzfJSv6M9T0
Here's the chapter for loop subdivision, although I suggest watching the whole video, since he doesn't really animate the loop subdivision part.
@theflyingpie For quads, we use Catmull-Clark subdivision.
With the way the old vertices are updated with local averaging, does the local averaging need to occur after all points on the mesh are loop subdivided before averaging is performed to prevent weird ordering dependencies on resultant geometry?
@weszhuang you need to use the original neighbor positions when updating old vertices, otherwise you'd end up over-averaging/smoothing. You subdivide first since we want to subdivide on the original, non-smoothed points.
I've been wondering why the update for an old vertex uses only its surrounding old vertices, as opposed to using the new vertices (which seems to make more sense to me because the update happens as a consequence of the split). But now I've realized the new vertices are completely determined by the old vertices, and that it would have the same effect to use old ones as using the new ones. The former is also more practical for computation.