Welcome to Fractal Forums

http://vimeo.com/65563772

I made this video a few years ago, but I didn't publish it as it had quite a few glitches in it, and I didn't have time to fix them all. However I think its still worth sharing as the glitches aren't noticeable, and it demonstrates the staggering complexity of the Mandelbrot set quite well.

It was created using perturbation theory (see http://www.fractalforums.com/announcements-and-news/superfractalthing-arbitrary-precision-mandelbrot-set-rendering-in-java/ for details), and actually I consider the glitches quite interesting as they show what happens when the perturbation theory goes slightly wrong.

Enjoy.

Final Coordinates:
s=1.440702e-209
real= 0.372137738770323258373356630885867793129508737741343871454145013316339415564311643633274959661448634647353127050570342 1437940529579667217009514171418668223134488754717790306510572753060176524459647591093062238923074187833545
imaginary=
-0.090398245434178161692952411151009819302665482564540025169572392202937018054290354292479312967254362819250116894458453910745765627096542515646237235211433447554951683193833635793401797382250559343517787867404041479514567004
Number of iterations: 2027420

Music:
Born Into It (chill out mix) by Redmann
Another Boring Lunchtime (edit) by Psychadelik Pedestrian
By The Water by Space Invaders v Psychadelik Pedestrian

I find it much too slow.

i checked your document, with the pertubation theory, sounds interesting, now i am going to check the "errors" :)

couldnt recognize any glitches, can you give an exact time location on where to find the glitches?

For example see the four yellowy green blobs at 12.07

is there a high quality version ? how long did this render?

I think it took about 9 or 10 hours to do all the calculations. The linked movie is at the full resolution but obviously has a lot of compression artefacts. I have a version that is 714mb that looks a fair bit better, but has no music.

such a small pixel resolution should be easy compressible, what codec you use ?
try h264 or xvid ....

That seems really fast indeed. What is the original resolution and which cpu did you use?

Its made from about 1300 images rendered at 1280 by 960. Each 640 by 480 movie frame is made from scaling and blending two of these images together. This means the centre of the image has a minimum of x4 super sampling, but it can be a bit less round the edges. It was done using an old intel i7 iMac.

I think you've done a fantastic discovery especially for mandelbrot set exploring. I can't wait until this is implemented somewhere so that I can use it. Unfortunately I can't program.

I just used Apple's iMovie as it was the only thing I could find that would add music, and then presumably Vimeo recompressed it using whatever they use.

vimeo is known for their high quality, so, i think the quality got lost before ...


either way ... the animation is far less interesting than if you would just provide us with a single image showing the particular error that occurs when using your described perturbation method, which at least seems to be veeeeeeeeeeeeeeery interesting!!!!!!!!!!!!!!!!!!!1one!!!11!

I'm not sure I understand, can't you use my java app?
http://www.superfractalthing.co.nf

Attached are two images to demonstrate the issue. The images are slightly different zoom levels causing the code to pick different reference points, causing one to work, and one to go wrong.
Also I have attached some text files which you can load into SuperFractalThing (http://www.superfractalthing.co.nf (http://www.superfractalthing.co.nf)) to generate the images yourself, although the colours will be different.

ah, i see, although still horribly low resolution, are you sure those blobs are not because to few iterations ?  inside these blobs you would need more iterations to complete the image...

I can use it, but there are bugs (it doesn't allow me to change the max iterations and coordinates, hangs often), and it lacks much functionality to be really useful. I think java also makes it slower than it could be by default, because it runs in a browser. Considering how fast it is already, this has much more potential than what we can do right now.

I had a look at them in the java application and I'm pretty sure it's not a lack of iterations. The max iterations is at 200000 while the green thing is just over 90000.

If you change the max iterations, then select 'refresh' from the menu bar it should work ok.
Do you have any reproduction steps to make it hang?

It hangs often when it finishes a render. This seems to happen mostly when I go close to the mandelbrot set (or a small mandelbrot set). The progress bar indicates that the render is done, but then the program hangs. With deep zooms, the program lags a lot but it still responds after waiting 10 seconds or so. The only way for me to go anywhere deep is to use the library.

At the moment it hangs so much that I can't even try your method to change the max iterations.

Did you try the low memory version? I suspect that the high memory version is exposing some stability problems in browsers. The high memory version seems to completely crash Chrome, but the low memory one seems to work ok. The two versions are exactly the same apart from the start up parameters.

I've made this for you
https://www.box.com/s/o6cob1yg1p5wolik0lcq
 :tease:

The blobs are caused by the points drifting too far away from the reference point, causing them to lose accuracy, which makes them indistinguishable from one another, and they come out the same colour.

right, there must have been some limit, dude, nice nice nice, this will improve fractal rendering very much!
thank you for sharing it with us, now i (or we?) just need to understand how to apply it to a rendering process, is your programm open source?