/sci/ - Science & Math

>desired roundness
Anywhere from 1 to infinity. Depending on the desired roundness

Yes. We'll take your variable roundness and then give you a constant answer which works for any roundness. Because it works like that.

>>3239803
you would need at least 12 polygons for something that could even in the loosest of interpretations be considered a sphere.

>>3239797 perfectly round sphere

Infinite polygons.
Even if you could manage that the renderer is going to draw the sphere using pixels, and pixels are square, so the rendering of the sphere will not be perfectly smooth. You need infinite pixels too.

>>3239810
not sure, even at 24 polygons in my software you can still see the edges.

op, i usually just put them at 90 - 360 polygons, but the more you add to them, the longer it takes to even interact with the object.

> 2011
> still rendering spheres as polygons

>>3239841
>CGI and not using polygons
>possibility

Pick one.

>>3239841
how do you NOT use polygons in renderers?

>>3239849
Give the GPU some basic information and then tell it to create and render the sphere itself, as big as it needs to. No need to polygonize it.

>>3239856
They're still polygons though, whatever mesh you're going to make still uses polygons regardless if it's GPU-cached or NURB'd.

>>3239861
You could render it as a point cloud of as much pixels as you need to fill the space.

>>3239866
But you're just using vertices, which is hardly useful outside of volumetric rendering depending on the intended goal for the point cloud.

Why not use equations to represent the shape instead of polygons? Like vector graphics, but 3d

>>3239849
VOXELS!

>>3239878
>Like vector graphics, but 3d
that would be nurbs generally, but expect to run into some frustrations if you plan to animate it look a person or creature

>>3239873
Why? It's not like a modern graphic card can't handle billions of voxels per frame. In exchange you get full roundness at every distance while still only rendering the parts you need to.

>>3239891
Because you're still not making a general shape to be used and rendered and properly. Also,
>while still only rendering the parts you need to
You're going to get the exact opposite result with a point cloud.

>>3239891
voxels are mostly beneficial in medical and scientific data visualizations, not for animation or rendering

>>3239905
oh, and volumetric displays too, but good luck having one of those on hand

>>3239897
> Because you're still not making a general shape to be used and rendered and properly
Huh? We're talking about rendering a sphere. It's not an 16th degree polynomic spline.

> You're going to get the exact opposite result with a point cloud.
You'll generate the points dynamically, depending on the view point. Not rocket science.

>>3239908
>You'll generate the points dynamically, depending on the view point.
No, you're going to generate all the voxels in a viewport regardless of what angle you're in, not like a filled object where only the face the camera is on gets rendered

>>3239908
>You'll generate the points dynamically, depending on the view point.
that's not what you get with a point cloud...

>>3239914
I know the view point used in the frame. I know the view port. I know the Z-buffer of the partial frame. How can I not generate only the points I need that will be seen?

>>3239925
How can you? It's a not solid shape or rendering, just represented pixel and geometric positions, you're only going to render all the ones currently present in that space, not just the ones you're looking at.

>>3239937
I do know the equation of the sphere and its gradient. Using that I can generate as much points as I need to fill a solid space.

>ITT: a perfectly round sphere doesn't exist in reality.. so why should it exist in CGI??!!

>>3240206
1). Because a CGI is using mathematic, i.e.: the closest it get to perfect roundness
2). lrn2hyperbole and semantics

>>3240212
The I in CGI stands for Image. That's something in the real world, not the imaginary world where perfect spheres can exist.

>>3240212
mathematics = theory
yes, it applies to the real world, but it's still at it's basis a theory.
also
with the usage of triangles, you'll never ever be able to achieve a perfect sphere,,

To get a perfect image, have at least four polygons per pixel (Nyquist Sampling in 2D)

>>3240263
>yes, it applies to the real world, but it's still at it's basis a theory.
>but it's still at it's basis a theory.
>a theory.

A geuss?

>>3240366
more of a proxy

Just ray-trace it in a fragment shader.

p = p0 + t*pv (parametric equation of a line)
p.p = 1 (implicit equation of a unit sphere)

Substitute the first into the second to get an implicit equation for t. Solve for t using quadratic formula, substitute into first equation to get coordinates.

Works just as well for cylinders and cones.

A torus is a total bitch, though. Quartic equation. Ferrari's method is unstable as fuck with single-precision floats.

>perfectly round sphere
>polygons

no way possible, surfaces are planar, so no matter the resolution(polys)..
it will never be a "perfectly round sphere"

>>3239844
ohai

Posting to dislodge religion thread from front page

This thread is full retard. You should all feel bad.

>>3242276
welcome to the club!

>>3242248
Holy shit, that depth of field was raped as hard as lens flares in a JJ Abrams movie.