Scatter On Point Clouds

[kodiak]

Hi folks,

What is the common way to scatter more points on another (lower rez particle - just to reason posting it in the FX section) point cloud in a way that it interpolates position (and maybe other attributes)?

I was into developing a similar renderman DSO couple of years ago that was used in a basic form on a project and gonna port it to mantra at some point, but a SOP based solution would suffice for now.

Creating a surface out of the lowrez one and scattering on that can work sometimes, but not a good approach in general.

thanks,

Andras

[Marc]

copy sphere's to the points and scatter on those? If you add noise to the spheres then it will vary the radius from the template point.

[Jason]

Similarly, you might be able to copy metaballs to the points, then with an IsoOffset SOP convert the metaballs to a Fog Volume, then use the Scatter SOP on the Volume. Now to interpolate attributes, use the AttribTransfer SOP against the original input points.

...

This does seem like a bit of a hack for what you really want: a thin-plate spline radial basis function interpolation - which really generates the best interpolated field for all the input points. This routine gets impossible to manage after a few thousand input points though, unless you have a kind of windowed system. I must say that I've always wanted a nice RBF interpolator like this.... perhaps Simon has some useful code in his Warp SOP development that could help with this?

[kodiak]

Thanks guys,

Yep, I had both approaches in mind but it was not extremely urgent I didn't really put the effort in there as they're more of quick hacks as Jason described (and my bigger issue was rendertime generation really, even though that wasn't mentioned in the post

Now, that I've finally finished on this project, I can move on and try to implement the mantra procedural (and maybe a SOP based version as well)..I might have questions later.

cheers

Andras

[sibarrick]

You could use my RBF morpher to sort of do this.... at least in a limited situation

You'll need the rbf morpher installed from here www.houdinitools.com

The attached was made in H8.2 - not sure if it works in H9. I locked the solution on frame 64 so you can see the idea even without the rbf morpher sop installed.

This relies on there being a constant number of particles to work. Notice how there are some gaps in the interpolated particles, not sure why that is....

rbf_particles.hipnc

[Jason]

I was wondering if it might be a cool idea to make a VEXdso from your RBF Morph SOP functions (and perhaps an expression) that can maybe return RBF interpolated attribute values at specified sample positions? Then perhaps one could generate interped fields as Volumes in H9 (using Image3D or expressions in a VolumeMix) - or do a straight query using points input into a VEX SOP or POPs or whatever. I can imagine sparse points with velocity vectors defining a vectorfield for particle simulations.

[sibarrick]

Interesting idea, the problem I see with this is that the RBF calculation requires 2 sets of points to drive it. The rest and deformed positions if you like. It doesn't just calculate a set of weights from the rest points it actually solves a massive simultaneous equation that maps one set of points to another, in the process you get a set of weights that can then be used to interpolate other positions. So in the case of a particle sim you'd need 2 sets of particles that map one space to another before you'd get any weights.

This is perfect for morphing but I'm not sure it works for particles. Or can you see situations where this would work? I can see it working in much the same way as my example in the previous post where you have a developing surface or an ocean or something but is that what you had in mind?