News: Visit the official fractalforums.com Youtube Channel
 
*
Welcome, Guest. Please login or register. April 24, 2014, 06:08:14 PM


Login with username, password and session length



Pages: [1]   Go Down
  Print  
Share this topic on DiggShare this topic on FacebookShare this topic on GoogleShare this topic on RedditShare this topic on StumbleUponShare this topic on Twitter
Author Topic: exponential-based complex dynamics  (Read 2485 times)
0 Members and 1 Guest are viewing this topic.
bh
Guest
« on: September 15, 2007, 08:08:37 PM »

Three images I submitted to the Fractal Art Contest 2007, but which didn't get much success there:


http://i16.tinypic.com/4uet7o2.jpg


http://i15.tinypic.com/6chc5ti.jpg


http://i16.tinypic.com/2ilfrr9.jpg

I am happy with the shapes I obtained. They have patterns characteristic of the dynamics of complex functions with exponentials. Such images are quite rare on the Web; I found only a few examples, like I think this image by Titia (very good work, by the way).

On a side note, I want to say that the contest organizers did a bad job compressing my fractals from the 1600x1600 images I provided. They applied the standard but sometimes terrible 2x2 chroma subsampling. Just compare:

http://www.fractalartcontests.com/2007/showentry.php?entryid=119
http://i4.tinypic.com/6d1e8lt.jpg
Logged
lycium
Fractal Iambus
***
Posts: 939



WWW
« Reply #1 on: September 16, 2007, 12:40:05 AM »

I am happy with the shapes I obtained. They have patterns characteristic of the dynamics of complex functions with exponentials. Such images are quite rare on the Web; I found only a few examples, like I think this image by Titia (very good work, by the way).

very nice renders! are you using a function like z <- a * e^(bi + c) + d, or something similar?

They applied the standard but sometimes terrible 2x2 chroma subsampling.

you're the first person i know of to also be aware of this issue; i am constantly amazed at how people are content to brutalise their images (especially deep reds) with untuned jpeg compression. in this respect i find irfanview much better than photoshop for doing such compression: you can disable the chroma subsampling in the saving options, something photoshop doesn't (afaik) let you do.
Logged

bh
Guest
« Reply #2 on: September 16, 2007, 05:06:29 PM »

very nice renders! are you using a function like z <- a * e^(bi + c) + d, or something similar?

It was:

f(z) =  z * exp((1-z^p)/(p*(1+c*z^p)))

with p = 3, and

c = 0.0539 + 0.1629*I
c = 0.0047 + 0.2087*I
c = 0.0139167 + 0.111517*I
Logged
lycium
Fractal Iambus
***
Posts: 939



WWW
« Reply #3 on: September 16, 2007, 11:12:09 PM »

thanks, i may have to try that one sometime soon smiley does it have a name?
Logged

gandreas
Explorer
****
Posts: 53



WWW
« Reply #4 on: September 17, 2007, 04:32:08 PM »

Especially nice considering how often the "fingers" get in the way of fractals like that (and if you increase the bailout to try to reduce that effect, you start getting moire effects, and eventually end up with Cantor dust since it's "fingers all the way down").

For a good introduction to the "BRD" set (and some other pretty pictures), check out:

"Growth in Complex Exponential Dynamics"
Logged
bh
Guest
« Reply #5 on: September 19, 2007, 11:23:56 PM »

thanks, i may have to try that one sometime soon smiley does it have a name?

It might be that I am the inventor of this particular formula smiley
Logged
bh
Guest
« Reply #6 on: September 19, 2007, 11:52:26 PM »

Especially nice considering how often the "fingers" get in the way of fractals like that (and if you increase the bailout to try to reduce that effect, you start getting moire effects, and eventually end up with Cantor dust since it's "fingers all the way down").

For a good introduction to the "BRD" set (and some other pretty pictures), check out:

"Growth in Complex Exponential Dynamics"


Hi Gandreas, thanks a lot for the link.

Actually, my formula is much friendlier than those in the article, like f(z) = lambda * exp(z). In my formula, if z is large, the formula is about like f(z) = z * exp(-1/(p*c)), which, assuming Re(c) > 0, makes z smaller; and if z is small, the formula is about like f(z) = z * exp(1/p), which makes z larger.

I said that my images are characteristic of the dynamics of functions with complex exponentials, but perhaps I should restrain my words. One thing I found characteristic of a certain class of formulas are the points where we see an infinite number of branches originating, and with a clear-cut separation along a line. These singularities appear because of the poles of the function w = (1-z^p)/(p*(1+c*z^p)); near the pole, w will be very large in absolute value, either with real part positive or negative, and thus exp(w) will be either very small or very large.

My images were still a challenge regarding anti-aliasing. I found that sometimes I needed 16x16 supersampling to get adequate results.
Logged
lycium
Fractal Iambus
***
Posts: 939



WWW
« Reply #7 on: September 20, 2007, 12:19:14 AM »

thanks, i may have to try that one sometime soon smiley does it have a name?

It might be that I am the inventor of this particular formula smiley

i guessed so; you're the best person to ask then right? wink
Logged

lkmitch
Iterator
*
Posts: 168



« Reply #8 on: September 20, 2007, 07:35:40 PM »

Such images are quite rare on the Web

Here's one of mine:

http://www.fractalus.com/kerry/gallery13/ladder.html

I think there are a couple of reasons why we don't see more exponential fractals:

  • The libraries for evaluating complex exponentials (including trig functions) are much slower than arithmetic computations
  • The standard bailout condition based on |z| doesn't usually work, particularly with z = c * exp(z).  Depending on the function, it can take some fiddling to determine the best bailout condition.

Both of these mean that exponential images can take orders of magnitude longer to compute than fractals based on polynomial functions. So maybe the tradeoff in taking more time is that using exponentials is one way to distinguish yourself as a fractal artist.

Kerry
Logged
lycium
Fractal Iambus
***
Posts: 939



WWW
« Reply #9 on: September 21, 2007, 01:08:41 AM »

  • The standard bailout condition based on |z| doesn't usually work, particularly with z = c * exp(z).  Depending on the function, it can take some fiddling to determine the best bailout condition.

i suppose one could check ln|z|, or equivalently check against a magnitude like exp(usual_bailout).
Logged

David Makin
Global Moderator
Fractal Senior
******
Posts: 2189



Makin' Magic Fractals
WWW
« Reply #10 on: September 21, 2007, 09:27:21 AM »

I think one of the reasons exp formulas are used less often is simply that there are fewer colourings that work well with them and you are more reliant on the shape of the actual fractal - a lot of folks seem happier playing with the colourings than with the main formula smiley
Same goes for the trig functions.
Logged

The meaning and purpose of life is to give life purpose and meaning.

http://www.fractalgallery.co.uk/
"Makin' Magic Music" on Jango
lkmitch
Iterator
*
Posts: 168



« Reply #11 on: September 21, 2007, 06:41:29 PM »

  • The standard bailout condition based on |z| doesn't usually work, particularly with z = c * exp(z).  Depending on the function, it can take some fiddling to determine the best bailout condition.

i suppose one could check ln|z|, or equivalently check against a magnitude like exp(usual_bailout).

Checking ln|z| would help with overflow, but the basic issue is that you can have spires and structures that extend to infinity, meaning that points in the same neighborhood can have orbits with arbitrarily large magnitudes and either diverge or stay in the set.  That's why a lot of colorings don't work for exponentials--most colorings are designed assuming that the orbit points are somewhat even distributed around the complex plane, or assuming that a large pixel value will lead to an orbit with large values.  And since, in the complex world, trigonometric functions are exponential functions, trig functions don't often work well, either.

Kerry
Logged
bh
Guest
« Reply #12 on: September 21, 2007, 09:15:41 PM »

I think one of the reasons exp formulas are used less often is simply that there are fewer colourings that work well with them and you are more reliant on the shape of the actual fractal - a lot of folks seem happier playing with the colourings than with the main formula smiley
Same goes for the trig functions.

Quite true.
Logged
lycium
Fractal Iambus
***
Posts: 939



WWW
« Reply #13 on: September 25, 2007, 09:37:02 AM »

the short answer is "not that i know of", but that has a lot to do with the nature of 3d vs 2d rendering...

one of the biggest differences between 3d and 2d fractals is that you no longer derive most of the beauty from colouring, but instead from structure[1], depending on how you render it.

for example, the recent apophysis 3d hack leverages the huge variety of fractals associated with flame rendering[2] and is reasonably flexible wrt colouring, however what you have in the end is nebulous clouds (unless you're really careful with your viewing angle etc). here's an example rendered with chaoscope: http://silent--dreamer.deviantart.com/art/Plasmic-Tornado-2-15520497

on the other hand, when you render solid 3d fractals, you get a real impression of depth with all the usual spatial cues (shadows, occlusion, ...), but you're more restricted in the sort of colouring you can do (usually) and even minor zooming can pose major technical difficulties. here's a recent example from a friend of mine: http://aexion.deviantart.com/art/Furnace-64157856

the video you linked was really interesting btw, and highlights some of these points well:

1. much research has been invested into creating that fractal!
2. it's a "once off" thing: it's a single, particular fractal, not a broad family. changing the colouring doesn't really change this (c.f. 2d orbit trap colourings).
3. the shading is a pretty typical early computer graphics model (diffuse / phong).
4. zooming is very limited (you eventually see spheres), and doesn't reveal much variety.

i suppose a lot of people are looking for the "holy grail" of 3d fractal rendering:

1. huge family of varied fractals, with intuitive controls for exploration.
2. flexible colouring / shading.
3. fast and realistic rendering, especially for zooms.

there aren't any renderers i'm aware of that can meet these brutal criteria, let alone fractal types! in summary: at the moment making 3d fractals requires quite a lot of specialist knowledge for a once-off result, unless you're willing to temper your ambitions: chaoscope, xenodream. this "fixed menu" has been around for some time (at least the 3d apophysis hack will inject some variety!), and all of these renderers have their shortcomings.

finally, i hope to attack this problem properly after my exams smiley i've got a pretty high-powered rendering system framework in place (effectively solving point 3 on the "holy grail" list), and have ideas for 1 and 2...

hmm that's a lot more than i planned to write tongue stuck out can you tell i'm trying to dodge studying? wink


[1] the structure vs colouring topic was recently brought up in this thread: http://www.fractalforums.com/index.php?topic=738
[2] by this i mean ifs rendering  la scott draves: http://flam3.com/flame.pdf
Logged

gandreas
Explorer
****
Posts: 53



WWW
« Reply #14 on: September 27, 2007, 12:04:37 AM »

for example, the recent apophysis 3d hack leverages the huge variety of fractals associated with flame rendering[2] and is reasonably flexible wrt colouring, however what you have in the end is nebulous clouds (unless you're really careful with your viewing angle etc). here's an example rendered with chaoscope: http://silent--dreamer.deviantart.com/art/Plasmic-Tornado-2-15520497

When I added 3D flames into quadrium | flame, the first thing I noticed was how cool things like a "julian badge" looked when extended to 3D as you twisted it a round or tweaked the parameters - right up until the point that you stopped manipulation the view space, at which point it turned flat.  If I made a movie of one spinning & changing, though, it was great.  Basically, the human visual system is really good at deducing 3D information from moving abstract points, but once those points stop moving, it falls flat.  For a similar effect, take a look at the movie at http://www.lifesci.sussex.ac.uk/home/George_Mather/Motion/BM.HTML - it's quite clear what you're seeing as long as it is moving, pause it and you just see a handful of dots.

Without some sort of depth cueing, non-animated 3D flame fractals are sadly somewhat less interesting than they really seem like they should be...
Logged
Pages: [1]   Go Down
  Print  
 
Jump to:  



Powered by MySQL Powered by PHP Powered by SMF 1.1.19 | SMF © 2013, Simple Machines

Valid XHTML 1.0! Valid CSS! Dilber MC Theme by HarzeM
Page created in 0.446 seconds with 29 queries. (Pretty URLs adds 0.019s, 2q)