Suggest a new filter.

New page! Continuing the story of dominos....

Just added a texture, also added sideways ones. Will need to figure the rest of the filter out over time.

Ok, honest opinion, image mixers are getting old. If we're doing image mixers, I'd like to see something even cooler than just throwing in images into the dots n domino pieces.

Now these are realistic looking dominos. Each domino brick is clearly outlined and divided into two halves. It's amazing how you got them to go horizontal and vertical.

Here's what you need to input: http://mrob.com/pub/comp/xmorphia/

lol

I'd love to see a FerroFluid filter that could create something like this, but I'm simply NOT the guy to create it. Anyone up for the challenge?


More info at this link:
http://www.geek.com/geek-pick/ferrofl...s-1500389/

Hi Erik, Regarding the ferrofluid see this link
http://www.filterforge.com/forum/read...sage119785

I think when it comes down to reaction-diffusion FilterForge just gets a bit slow on it all. I suggest doing it manually via actions in Photoshop.

I made a video tutorial you can watch here: https://vimeo.com/61154654

Erik you might like this. it's buried in the thread I mentioned.

Reaction_Difussion with Images.ffxml

Sorta sloppy attmpt at it...

Niiiiice, ThreeDee! I like what you've got so far!

Plexus-Nexus! Plexus ROCKS - WOOHOO!!!!!!

An interesting challenge...

Hey ThreeDee, Were you by any chance going to post the PlexusBlur to the library??? I'd love to play around with it! You do such amazing stuff.

Did some more work on it today, not ready to submit.

Added Sphinx's "true variable blur" effect among other things.

Really cool effect!

Yayyy! THANKS for this!

I second this! Complexification can create some incredibly beautiful patterns!

Here's another cool idea http://www.artfromcode.com/?p=853 regarding Lichtenberg figures

The author was nice enough to find an old code sample though I'm not sure if that can used as an idea by the scripters here http://jsbin.com/yufubege/5/edit

This basic branching pattern is found everywhere in nature from rivers, erosion, plants - and a filter or component like this could be useful.

That is very cool @mitaywalle! Playing with it now.

cool Rick, slowness was main drawback for me so when i use it I would have to generate small then vectorize it to enlarge. curious to see what you come up with.

One last image from my reaction diffusion experiments. Almost done with the filter, this one took 80 seconds to generate (600x600), not bad for all the blurs and high pass calls. This is probably one of the presets. Thanks again to SpaceRay & Skybase for the original reaction diffusion code.

Yes. The original had 7 passes (High pass, Threshold, Blur). I left the first 3 passes alone. On the 4th pass I replaced the blur with "edge protected" noise, the 5th pass is just a single high pass with the contrast turned up. In the end I used 5 High pass, 4 Threshold, and 3 Blurs in the reaction diffusion part of the code.

The speed of a full render is very size dependent. Previews (and 600x600) are not too bad (all under 2 minutes) and FF caching speeds up some previews to under a minute.

I am on the fence about releasing it. Large renders (over 2000 pixels) take quite a while if you have less than 4 cores in your CPU. I am doing more testing on my slow & fast CPUs today.

Should I release it with a warning about the speed on older (1 or 2 core) CPUs??

I think its where it gets me. Basically if the number of iterations is too low, then the reaction diffusion doesn't occur on some images (specifically where images have low contrast), hence to keep the output uniform it needed an extra number of iterators. It's also the case where it kinda takes 20 or 30 more iterations before "neat patterns" emerge.

There's also the case that FF's pretty sluggish when it comes to this which is why I also did reduce the number of iterations on purpose. In doing so, I set the highpass radius larger to increase the reaction speed but in doing so, it sacrifices output quality and purpose. In the end it kinda results in a filter "that can be achieved through faster means". I can basically achieve a similar appearance via some other combo of components if it's at lower iterations. What makes "Reaction-Diffusion" unique is the fact that as the iteration count increases, more unique patterns emerge not because of its results.

At the time I uploaded the FFXML files there were enough people who were screaming "FF IS TOO SLOW!!" blah blah blah and that basically made me avoid uploading any filter that took around 1 minute to render. So to me what I've uploaded to the forums should be considered merely an example / education material/ or an idea utilizing iterators as a way of achieving some level of detail. I personally feel reaction-diffusion as a filter doesn't suffice what I consider "quality".

But all aside, just note that this is all my super strong personal opinion about stuff. I'm merely basing this point on reasoning alone! In the end it's up to you. If you think it reaches a certain quality, I think it should be part of the FF library.