CorvusCroax's Junkyard

Quote
StevieJ wrote: Yes, if you could create the fractal patterns of bling 1-7 in FF without inputting them.....which would be some complicated curve work....

Yeah, that's a really interesting image. The smooth swirls look really great when converted into a height map; they sort of etch in and out in a really pleasing way. It's gotten me wondering how to make it. Illustrator would be my first choice. However, it uses a certain amount of reflection and self-similarity. I think you could do something similar with some inputs, but I think it would be really hard to do it in any controlled way. Maybe with Processing ™.

Quote
StevieJ wrote: ...but users would eat up the "ornamental" value in wood, cement, and metal...

Well, it's submitted.... I guess we'll see.

Yes, indeed. FF might also benefit from a low complexity fractal flame node, too.

I really wish that FF had 'bomber' support, using both raster images and SVG (or even TTF) source inputs. To this day, I STILL dust off KPT, just for the KPT Scatter function.... even though it's like a decade old now. A simple implementation of something like KPT Scatter would be great inside FF. Imagine an FF filter where you could give it a series of images, and it places them at designated spaces inside a FF green node, which you could designate proper rotation, scaling, tint, etc. Now imagine that instead of a basic raster image, those scattered objects be channel inputs... we could make incredibly wild looking constructions inside FF.

Btw: has anyone used Aviary ™? In particular, Aviary Peacock http://aviary.com/tools/peacock It's another node based image generation program, but seems to be more like illustrator, but I really haven't looked at it that closely.

Quote
StevieJ wrote: Vlad stated that bomber components are supposedly coming in V2.....

Oooh... really? Sweet! Do you have a link where they're talking about it?

thanks for posting those originals, corvus

i like those last images, too. especially the first of the last, the lily. see, i told you you shld have called this filter 'gilding the lily'

Quote
Kraellin wrote: i like those last images, too. especially the first of the last, the lily. see, i told you you shld have called this filter 'gilding the lily'

Yeah, but it sounds dirty when I say it.

Quote

CorvusCroax wrote: I wish there was a way to extract the dominant bright color from an image. Maybe like a way to make more selective the 'average color' component to particular parts, as defined by a mask. (you could use HSB extract to define the mask). So, for example, in the image above, I'd like to pick out that big green-yellow really saturated color under the bridge arch, and then put it back as a color input. Does anyone know a way to do that? (maybe some clever trick with the max levels component that no-one seems to use?)

A number of ways you could do that. The simplest way would be to do something along the line of two sliders which are the x and y coordinates for the color to be picked. Like Mike Blackney's Pixel Color. You could also blur the image a bit to get an "average" or a few pixels.

One good way to get a flat color output extracted from an image is to plug an image into the Elevation Gradient component. Use a flat black as the "Elevation" and an image with the desired color in the top left corner: you get a flat fill of that color.

The other way, closer to what you are describing, would be to do a simplification of the "Color Palette from Image" filter I made recently, and just pick one color, by hue. (This involves motion blurs etc. to get flat colors.)

Hey, here are some new pics of a thing I've been working on: the Rust-erizer. Takes an alpha channel, and renders it as rusty metal. I started with Aurum's manhole cover filter as a base. Here's my Standard test object:

rusterizer 003: again, a standard heightmap test object. I think it needs to fill in the gaps more.

Rusterizer 009: machines! You may recognize this heightmap from the 'redball' post previously. Dang, should do more of those. Looks HEAVY

Rusterizer 012: more machines. This is really creepy, actually. Put some 'melt' into it, which uses a blur. Really tears it up.

Rusterizer 015: with yet another height map. Thinking this might be fun for a steampunk machine - maker filter.

Rusterizer 017: source image is an xray of some kind of electronic circuitboard thingamajig, a hard drive, maybe. Have added a control to allow for mixing of the rust intensity in 3 different ways: uniformly, with a balance control, using some perlin noise (again, using the same balance control) or using the input image selection; that way the user can just go in and manually choose where to have more or less rust.

Rusterizer 020: another sample object. It looks pretty good with a fairly flat source. These are just overlapping boxes, with just a tiny bit of relief. Again, thinking this might be good for a kind of steampunk rusty wall panel maker if it was just more visually interesting.

Rusterizer 022: using some test text.

Rusterizer 029: again, a standard test object. There is a tiny tiny bit of blur which helps smooth out the 'stair step' artifacts of the input image.

Rusterizer 033: this one is a bit complicated: the paper is just some paper. The background swirly bits are a highres photo of some smoke, which has been RUSTerized. Then the little foreground creatures are Haeckel's drawings of plankton. They look quite strange all RUST-ified.

nice. but i hope there's a brightness control. all but the last one are pretty dark.

Thanks for the encouragement!

Quote
nice. but i hope there's a brightness control. all but the last one are pretty dark.

Yeah, there's a brightness control. I think that actually it had more to do w/ me working on these late at night, with most of the lights off. Here they are w/ more normal brightness:

#13 with the rust in the cavities. The cavities are made simply by using a high pass.

oh yeah, much better

So, I've been trying to make a good cutline filter for a while. I ended up making something else instead: more like an improved Comic Book filter. As cool as Comic Book is, it's not very clean. If you use more than 4 steps, skin tones go all weird, and you kind of get what you get, based on the average.

So, I made one which allows you to specify exactly which colors you want to preserve from the image. It works pretty well. Check it out (links below)

The end result reminds me of Shepherd Fairey! you know, the guy behind the 'Obey the Giant' and 'Andre the Giant has a Posse' phenomenon. Oh, he also is the guy behind this artwork, which you've probably seen)
http://obeygiant.com/archives

Here's another:

#4, the original