Fun With Cookies

[raytheon]

hi all..

I'm trying to create a smooth mesh from the union of a box and sphere with high detail in areas where there are sharp changes in topology. I've used the cookie SOP to do the union. If i pipe it into the Convert SOP it generates a bad quad mesh. If, instead, i pipe to an ISO Offset SOP to convert to an isosurface using ray intersect then the mesh is cleaner, but it lacks refinement in areas of high detail and is too dense in areas of low detail. The surface of the sphere is also quite angular even if the uniform sampling is tripled from the default value.

The Cookie SOP takes the polygonal geometry as input and does a boolean union on them. What is the Cookie SOP doing and what does it produce?

The Convert SOP output improves significantly if the "Convert To" is changed from "Mesh" to "Polygon". It seems that the operator generates bad topology when creating the quads. The mesh in either case still contains obvious polygons with more than four sides. Does anyone know what is happening here?

Is the ISO Offset SOP converting the polygons to voxels internally? If so, how is it generating the isosurface? Marching cubes, perhaps? Is there something like the ISO Offset SOP which generates the mesh adaptively so that quad/tri density is highest in areas of greater topological complexity?

Thanks.. i'm new to Houdini so it's probably a FAQ :>(.

Just a note - in the attached .hipnc please change the type from Crease to Boolean in the Cookie SOP

FunWithCookies.hipnc

Edited March 7, 2010 by raytheon

[raytheon]

FYI a reply from the sidefx forum says that Houdini does not provide this feature out-of-the-box.

Out of curiosity, the Convert SOP appears to be converting each primitive to a separate triangle mesh. I'm pretty sure this is a FAQ, but is there a good way to create a single triangle mesh over all primitives?

Edited March 10, 2010 by raytheon

[edward]

TriStrip will reduce the number primitives.

[raytheon]

Yes.. i see that it reduces the 80 polygons of the sphere to 13 triangle strips, but unfortunately http://www.sidefx.com/docs/houdini9.1/nodes/sop/tristrip states that triangle strips aren't usable by most SOPs and mantra won't handle them. The same source says it is possible to convert tristrips back to triangles with the Convert SOP, but if i try using

From Type: Triangle Strip

Convert Type: Mesh

Connectivity: Triangles

it still outputs 13 triangle strips so i'm probably doing something wrong or my expectations are optimistic. Is it reasonable to assume that this 80 polygon sphere can be converted to a single triangle or quad mesh?

[edward]

Mantra renders tristrips fine as of H10.0 (?) or so. The TriStrip algorithm is a greedy one and won't necessarily find the most optimal one in general. I don't have time to look at your file. But it also might be impossible to represent as a single tristrip. As for "mesh" primitives, I believe they must have a quad-like connectivity. I highly doubt any triangulated sphere will be able to do that without holes at the top/bottom caps.

[raytheon]

Yes - i agree that i was being optimistic. The TriStrip operator seems to work as advertised on my model, though the resulting triangles are, as you say, not optimal. I have a more complex scene. If i pass the geometry through PolyReduce then TriStrip it cleans up the triangles a bit. I was curious how other companies handle it and found Joao Pedro Jorge's 2008 Masters thesis at http://www.jpjorge.com/index.php?headfile=portfolio.php&topic=3. In case you or anyone else is interested (probably old news) it describes some Houdini SOPs he wrote for Digital Domain to do procedural mesh improvement.

thanks once again for your help!

Edited March 19, 2010 by raytheon

[edward]

Thanks. I hadn't seen that link. I'm not sure why he needed to implement a Houdini SOP for "Mesh Simplification using Edge Collapse with Quadric Error Metrics" though. I think that's just the PolyReduce SOP.

[raytheon]

He does mention the PolyReduce SOP in section 3.4.. apparently it wasn't flexible enough for their needs at the time.