Laplacian fixer ICE

Discussions about SOFTIMAGEs© Interactive Creative Environment©
User avatar
rray
Moderator
Posts: 1810
Joined: 26 Sep 2009, 13:51
Location: Bonn, Germany

Laplacian fixer ICE

Post by rray » 27 Nov 2015, 01:29

Hi, I was planing to come back a little to modeling, and do think this tool by Hans Godard would be so helpful in a lot of situations (see 5:09 mark in this video)



Does anybody have an tip on how to go get started writing something like this, ideally in ICE?
It's like a smooth (vertex average) but it fits much better into the unselected verts.
no idea how to approach that technically.

-rr
softimage resources section updated Sep 26th 2026

User avatar
Tekano
Posts: 488
Joined: 09 Jun 2009, 12:49
Location: London, UK

Re: Laplacian fixer ICE

Post by Tekano » 27 Nov 2015, 09:14

Hi rray here is the algorithm, looks pretty easy to do in ICE, not sure about the plugin if you get stuck can knock up a quick tree, it looks like its just move each point to nearest 4 neighbors or very similar


https://en.wikipedia.org/wiki/Laplacian_smoothing
Gossip is what no one claims to like, but everybody enjoys.

User avatar
Mathaeus
Posts: 1778
Joined: 08 Jun 2009, 19:11
Location: Zagreb, Croatia

Re: Laplacian fixer ICE

Post by Mathaeus » 27 Nov 2015, 09:39

From video, looks like proportional tweak tool performed after tweak, bounded by two selections. With only one direction of shrink or grow, according to 'details are coming from' option. I think there is no average of point positions, directly. More like average/smooth of some scalar attribute, used to modulate the smooth step (or something like) interpolation of delta against original. While 'repeat smooth' method is just stopped once the bound (defined by second selection) is reached. So in ICE world, something similar to ''deformation with bulge'', or "dent and bulge" factory sample - but, with plain 'repeat smooth' instead of 3d distance based one. It seems he used similar method in other parts of video, too.

User avatar
rray
Moderator
Posts: 1810
Joined: 26 Sep 2009, 13:51
Location: Bonn, Germany

Re: Laplacian fixer ICE

Post by rray » 27 Nov 2015, 14:22

Tekano thanks for the link. This is probably used somewhere as a part of it but I think there must me something more going on than just point smoothing, some additional step, because the transition into the unselected points is so nice and continuous. I tried but never could get this effect using just smoothing.

Mathaeus you mentioned deltas which made me think about that it might work similar to delta mush. Going to read up on how that works again, maybe it's the deltas that are smoothed with the Laplacian method.

Cheers
-rr
softimage resources section updated Sep 26th 2026

User avatar
rray
Moderator
Posts: 1810
Joined: 26 Sep 2009, 13:51
Location: Bonn, Germany

Re: Laplacian fixer ICE

Post by rray » 27 Nov 2015, 15:07

I'm wondering whether it works with only two selections, or maybe any number. If he uses something like shrink selection, maybe there's some way of "bleeding" the unselected transforms into the selected parts.
softimage resources section updated Sep 26th 2026

User avatar
Mathaeus
Posts: 1778
Joined: 08 Jun 2009, 19:11
Location: Zagreb, Croatia

Re: Laplacian fixer ICE

Post by Mathaeus » 27 Nov 2015, 20:21

rray wrote:I'm wondering whether it works with only two selections, or maybe any number.
At least, my explanation is based on two selections, where one is start, second is end of effect. According to last example, with mouth, where effect does not extend into complete selection - I think there's a kind of grow from first selection to the just a first bound of second, where steps were used to modulate the final offset.
According to first example with cylindrical thing, where it takes the modified orientation into account, too - I'd believe this is blend of two complete transforms, created by movement of manipulator (smooth step, bezier, whatever), modulated by mentioned grow-shrink. "Grow-shrink" is "smoothed", here.

User avatar
FXDude
Posts: 1145
Joined: 19 Jun 2012, 19:59

Re: Laplacian fixer ICE

Post by FXDude » 28 Nov 2015, 19:11

That clip shows awesome work by the way.

Yet all require python and or C++ to acheive. Whereas in soft...