Fractal Animation

[ianc10]

Hello..  here is a link to a video I made of zooming into the set (note that it requires a DivX http://www.divx.com compatible decoder to play it):

http://www.mandelmania.net/mandanim.avi

I'd be pleased to hear what you think of it.

Thanks,
Ian Calegory

[Jules Ruis]

It takes one minute to download it (20 MB).

[Jules Ruis]

I donloaded the file for the Windows Media player, but I hear only music, the images stop to continu after intro,

[ianc10]

Do you have the DivX encoder installed?  You can download it from http://www.divx.com

Ian

[Zoom]

I like your efforts, but I hope you won't mind a little constructive criticism.

1. You have a lovely 720x480 video, why put the render at less than that? Unless it's an artistic decision to have the black border surrounding the animation.

2. Your decision to color the interior of the Set is one not taken by most fractal animators. Actually I consider it very brave of you, but you're bound to get complaints from someone. Never mind those complaints, you did a good job with that.

3. This work is one of those "infinite" zooms, as all you do once you get to the attractor point is zoom in on it. This is great, but the movie doesn't have a running time of forever. What some people do is focus towards the end on one of the points near the attractor, which ends up as a mini-Mandelbrot. I prefer that method myself, but there are other things.

4. Obviously there is a problem with your calculations toward the end. You seem to be running up against the limits of precision. What's interesting here is that you make it the conclusion of the video, as if it were a special effect. Again, I wouldn't do it but that's because I don't know if I could do it as Ultra Fractal 4 goes very deep indeed. It's interesting.

5. Finally, the image is rather muddy, like a JPG flipbook. Sadly this happens to a lot of us-- YouTube shrinks and muddies my work, for example. But yours is coming straight from your own website, so I wouldn't know what the trouble is.

I would also point out that of the prints you're offering on your site, the third one has a humongous hole in it that needs more iterations. Also your wallpapers really lack decent anti-aliasing.

I think I'm done. Sorry if I killed you.

[ianc10]

Helloo...  ya that faded out the video where the resolution limit is reached.  It's the double data type's resolution limit.  I pretty much hand-coded everything (except that I used a movie editing program to compile the frames into an animation).  I'm working on another video that goes somewhere besides an infinite attractor, and it's a lot more interesting to watch.

The choppy video is because I just basically threw it together with DivX...  I'm planning on fine-tuning the encoding process, and using something other than DivX--probably just an MPEG encoder.

You mention anti-aliasing...  I have thought of ways to anti-alias the images and haven't done much with that.  Are you aware of a good anti-aliasing algorithm to use?

Thanks for the pointers,
Ian

[lycium]

You mention anti-aliasing...  I have thought of ways to anti-alias the images and haven't done much with that.  Are you aware of a good anti-aliasing algorithm to use?

good antialiasing of fractals is actually incredibly difficult. the trouble is that they (theoretically) have infinite frequency content, which does not bode well for any amount of supersampling (thanks to nyquist). the ideal way to antialias the image is to "bandlimit" it, which basically means cutting off all frequencies after a certain point the nyquist limit. this is very difficult to do analytically/perfectly, so in practise a great deal of empirical tuning needs to be done; for example, just limiting the number of iterations produces a smoother signal, and the question becomes "how many iterations should i use given a specific zoom depth?" the whole thing needs a ton of tweaking if you're going to go down that road and try to do it optimally.

otherwise, just straight supersampling works pretty well :) it's of course best to do this in the code itself rather than via large images (speed, quality, storage), but simply rendering the images at some multiple of your final resolution (2-3x in each dimension for a start) and then doing a high quality resample to the final frame size makes a world of difference already. i recommend virtualdub for this - its resizing filters are excellent.