Ooh-videogames
Regular
I was checking out a Nintendo patent on the use of Polyhedrons, so I'm just looking for a better understanding of the term.
Polyhedron, is it a different form of polygon use?
- Thread starter Ooh-videogames
- Start date
Ooh-videogames said:
Huh? Polyhedrons are just solids are far as i know. They are made of polygons. Don't think Nintendo patented the use of solids in games or that would be a lot of licensing fees for them since the first polygonal games were created.
No. You need to model in them from the start because it is extremely difficult to subdivide, an arbitrary polygonal model, say of a duck, into an efficient set of convex polyhedra.PC-Engine said:
Believe me - considerable effort was spent investigating this. Certainly, once you have convex objects there are many benefits, but getting there is tricky.
By Polyhedra do people mean CSGs? 3D volumes with surfaces? Or non-triangular surfaces? Otherwise it just sounds like closed 3D triangle meshes to me - any polyhedra such as a dodecahedron can be modelled and rendered already.
What? No PVR message icon?
This system modelled convex polyhedra (potentially with missing faces and/or unbounded faces), from a set of infinite planes. There was (usually) one plane per face of the polyhedron and each plane would require approximately the same data as a single triangle. In effect, a polyhedron was constructed using a CSG "intersection" operation.
Well, I suppose it could mean a multitude of things, but what I was describing was the system used in PowerVR series 1.Shifty Geezer said:
This system modelled convex polyhedra (potentially with missing faces and/or unbounded faces), from a set of infinite planes. There was (usually) one plane per face of the polyhedron and each plane would require approximately the same data as a single triangle. In effect, a polyhedron was constructed using a CSG "intersection" operation.
Objects built as solid 3D mathematical volume representations (CSGs) can be rendered extremely fast. One example that comes to mind was the raytracer ralled 'Real3D' that appeared on the Amiga in the 80s. It worked almost entirely with CSGs, building up objects from combining primitives such as cylinders and spheres, and rendered scenes 2-5x faster than I think every other application. More complex operations are achieved with Boolean combinations. eg. A large sphere has three small cylinders intersecting it's surface with a Boolean 'Subtraction' applied to poduce a bowling ball.
Where speed and accuracy is a definite plus for CSGs, their flexibilty is limited. It's not too easy to model a realistic human character with cuboids, spheres and cones! Whereas for something like a Snooker Table they might be quite useful, any hardware optimizations for CSGs will see limited use, so it'd make more sense to use that silicon in mesh rendering and use meshes to represent what could otherwise be more effectively represented as CSGs.
Where speed and accuracy is a definite plus for CSGs, their flexibilty is limited. It's not too easy to model a realistic human character with cuboids, spheres and cones! Whereas for something like a Snooker Table they might be quite useful, any hardware optimizations for CSGs will see limited use, so it'd make more sense to use that silicon in mesh rendering and use meshes to represent what could otherwise be more effectively represented as CSGs.
I can this method used along side laser scanning like they do with CGI models. For example 3D laser scanning a miniature sculpture would be much better than building the 3D model from scratch. Also since CSG is used in CAD a lot of man made objects could be ported very easily since it's all CAD data.
They've even managed scanning models from a camera. I think there's even EyeToy scanning. (http://www.gamesindustry.biz/content_page.php?section_name=dev&aid=2860). These build meshes AFAIK. Working out a mathematical volumetric model of a human head capable of portraying subtle muscular and bone variations from a standard would be...an incredible feat.
Shifty Geezer said:
Using a camera will get you nowhere near the accuracy of laser scanning, that's why maquettes are scanned with laser for CGI models used in movies. CAD uses CSG data. CNC milling machines also use CAD data. These are the machines that carve scale the miniatures into large models used for animatronics.
Last edited by a moderator:
Realsoft is still around: http://www.realsoft.com/Shifty Geezer said:
Maybe offset mapping can be used to give a bit variation on the geomtric shapes?
Anyway, CSGs could run on a CPU optimised for the purpose, and stuff not suited for CSG could go to the GPU.
Ooh-videogames
Regular
london-boy said:
Oh, I wasn't thinking that. I just wanted to know about polyhedrons. Thats why I didn't ask why Nintendo would patent the use of them.
Last edited by a moderator:
Companies have a habit of applying for patents no matter what, even when it's known and well established technologies, and the US patent office has a habit of granting patents no matter what, even for known and well established technologies.
Similar threads
