ompf logo

(L) [2008/04/22] [beason] [ompf logo] Wayback!

I converted the ompf logo to a 3D model:
ompf.png
I also played around and created an IFS to make the attached image below (and a [LINK http://kevinbeason.com/ompf/ompfFlake.png 1920x1200 PNG version]), using 14 iterations.
This zip file contains both the original model and another model containing 6 iterations of the IFS, plus a text file containing the IFS transformations.
The models are in Open Inventor (.iv) format. You can convert to other formats by importing/exporting in Blender.
The ompf logo has 1756 triangles. The iterated version contains 3906 instances (6,858,936 tris).
ompf.zip

(L) [2009/08/27] [Shadow007] [ompf logo] Wayback!

I thought about some way to do an OMPF logo with the four (a little bit more rounded) letters each having different materials : one completely reflective, an other quite refractive one, and the other with different colors/reflectivity....
To make sure there is something to reflect/refract, the letters would have been at different depths. (with their sizes adjusted so that they would appear homogeneous).
And finally perhaps all of this put inside of a Cornell Box ...
I absolutely don't know what it would give after a render, but if someone has the time to try [SMILEY :)]

(L) [2009/08/28] [tbp] [ompf logo] Wayback!

Sadly that ompflake isn't as convenient as the good old sphereflake to (en)code, but perhaps someone can come up with a trick that escapes me at the time of writing?
It's tempting to learn flash just to annoy everyone around here with an interactive rendition on every page [SMILEY ;)]
Anyway, it's superb, even if i'm not quite sure that pink/purple is manly enough.

(L) [2009/08/28] [beason] [ompf logo] Wayback!

>> tbp wrote:Sadly that ompflake isn't as convenient as the good old sphereflake to (en)code, but perhaps someone can come up with a trick that escapes me at the time of writing?
The technique is exactly the same as the implicit sphereflake, but substitute sphere with (ompf) model. Each level of the (implicit) hierarchy is just a bounding sphere which is big enough to contain all possible children. What it actually contains are 5 bounding spheres, the children (each like the first but with their own transformation), and each is ray-tested in turn. The tree is traversed up to a max depth. The calculation of the "big enough" radius is remarkably straightforward, and is laid out along with everything else in:
Efficient antialiased rendering of 3-D linear fractals
by John C. Hart
[LINK http://graphics.cs.uiuc.edu/~jch/papers/]
(see section 3)
I found the paper after ogling at lycium's piece here: [LINK http://lyc.deviantart.com/art/rise-64751071]
so props to him. I also heard he was gonna make an ompf logo so perhaps I am guilty of stealing that idea as well.
It was a good idea so if I stole it I apologize, lyc!

(L) [2009/10/13] [phkahler] [ompf logo] Wayback!

It's not really an IFS. If it was, you would not need to define the Ompf shape at all, only the transformations. I am of course referring to IFS as defined by Barnsley in "Fractals Everywhere" using the collage theorem. Is there another "IFS" that I'm not aware of?
You did make a nice recursive thing though. Like it.

(L) [2009/10/13] [beason] [ompf logo] Wayback!

I think John C. Hart calls this kind of fractal a "linear fractal"... except mine includes the intermediate levels of recursion instead of just the limit. Also instead of a sphere I use a 3D model. If I misused "IFS" I apologize, it's the same to me: one just takes a (set of) transformation(s) and applies repeatedly. I don't have a rigorous understanding of fractals or even know the collage theorem.
Thank you for the compliment though [SMILEY :)]

(L) [2009/10/15] [phkahler] [ompf logo] Wayback!

Some explanation:
For an IFS, you take the whole image and completely cover it with smaller copies of itself - i.e. make a collage out of it. You then write down the transformations that take the whole and map them to each of these smaller copies. To render, you just start with some random point and apply the transformations at random (possibly weighted by their area relative to the whole). This is the "iteration" of the "functions". The sequence of points you get converges to an attractor shaped like the original image (thats the result of the collage theorem). If you don't cover the original completely, you'll get a fractal that looks similar to it.
For your Ompf logo, you provided some of the transforms - the ones that take the base Ompf to the smaller ones. In fact I think your transforms take the *whole* and map it to the smaller ones. The part that's missing is to define transforms that take the whole and map it to the base Ompf - and that would take a lot especially in 3D. However at that point, you could dump the model and use just the transforms. I suppose you'd render a sphere at each point visited by the IFS.
I haven't seen many 3D IFS fractals - probably because it's too hard to do the collage [SMILEY ;-)]
On a related note, I like to ray trace a lorenz attractor with spheres in real time. I just add spheres and delete the tail after a few thousand. It's a nice test scene that doesn't require loading any geometry [SMILEY ;-)]

ompf logo back

Board: Board index ‹ Ray tracing ‹ Data sets