fractalius -PS Filter effect

We're getting closer with Beta 2.0. But the way those lines follow the curves so nicely hasn't been achieved yet.

My attempt:

I read in another forum that the technique employed is called "line integral convolution" and here is website showing the math behind it. Maybe this helps to getting another step closer

okay I give up for the night but just had a bash at a filter not there yet but this is image 1

Quote
"I read in another forum that the technique employed is called "line integral convolution" and here is website showing the math behind it. Maybe this helps to getting another step closer"

I am just posting this again, because I saw I forgot to send the link I mentioned. Or is it forbidden to post external links in here? http://www.zhanpingliu.org/Research/F...IC/LIC.htm

Quote
Kraellin wrote: wow, i like that, bel! it's like a hyper-sharpened cat face. that could be useful!

hey it was 2 am.. and kitty pic was easier to find.. mind not sure what it will look like with Topaz my black cat.

plus got a ways to go trying to figure out additional bits to it (or turn it into a totally differen filter..um).

Quote
MetaVampire wrote: I am just posting this again, because I saw I forgot to send the link I mentioned. Or is it forbidden to post external links in here? http://www.zhanpingliu.org/Research/F...IC/LIC.htm

Interesting.. I love some of the images on the home page rest I would need to plug my brain in for...

Quote
StevieJ wrote: I think that's over-complicating a simple effect. You can easily emulate it.....starting simple by combining/modding these two filters.....

both cool filters but the first looks closer to the fractulus effect.

Oops! Just realized that the Line Integral Convolution filter I included in the last message I made in the 2.0 beta. Sorry. Won't work in 1.0. Didn't use anything 2.0 specific, but just happened to be in 2.0 when I did it. I'll have to be more careful in the future.

Thanks Ben. Here's a better one. The original just did the "Manhattan metric" for coloring by adding the two noises together. This one calculates the actual distance. Well, approximately, anyway. I could get dx^2 + dy^2 but then there is no function for square root so I used Gamma and made sure it was 1/sqrt(2) at 0.5 so it's a reasonable approximation. I still give a checkbox for the manhattan if you like. I also took the image out as a source. Easy enough to put back in, but I didn't want it turning up when I hit the variants key.

I'd like to do coloring for angle also but there's no easy way to get that from (dx,dy). I thought of using gain as an approximation to an inverse sine but that's not very pretty (or accurate I think). I think I have a way now, though. I can't get the angle from (dx, dy) but I can get (dx, dy) from the angle by using the sine function so I think I can use a noise function to signify the angle, convert to dx, dy and then a separate noise for the distance which I'll combine with (dx, dy) using a multiply blend. That way I can get the distance and angle exactly and be able to color based on either one. Really, I think in the end that the distance is more aesthetically pleasing since it is based on the local "look" of the texture rather than just what angle it's headed at so maybe I'll do the angle and throw it all away when I'm done. Haven't tried it yet, though, so maybe I'm wrong.

Oh, and here's a big thumbs up for any more sophisticated math/programming constructs like the loop you mention. I wanted to change the "kernel" in this convolution which weights the various distortions, but I can't really do that either since it's embodied in the remapping into those distortions and you can't change those remappings at runtime. Another argument for my post elsewhere about more flexible remappers and/or nodes which take in a numeric argument and apply a function to it with a numerical result.

Anyway, thanks for the kind words!

Darrell

LIC.ffxml

looks good, submitting it?

@Carl - yes, I have submitted the filter, now called fractalitious - my first one!

@CorvusCroax - Thanks for the snippet, it is has some similarities to the motion blur/sharpen/blend technique in this filter, but has given me a few ideas to try out on an upgrade to this filter.

and another...

and a last one showing distortion madness...

Very cool. Please submit.