Multiple-image input workaround

sounds like jffe's mirror filter would be handy here.

Here we go. Four input images plus four alpha channels...

To make it possible to use several images, they have been pre-composed in Photoshop onto a two-by-two grid.

As the RGB image, we have here a composite of four images: Background image, reflection image, diffuse map and a lathe profile.



As the alpha, we have a composite of bump map, reflectivity map, transparency map and metallic map.



Then we load alpha as the current selection and run it through a specifically constructed FF filter which uses all 8 images to compose the following output:



Well, there you have it, FF with 8 custom images; background, reflection, lathed profile, color map, bump, reflectivity, transparency and metallicity -- all at once. And, since no HDRI mapping is used, this is even a "simple filter," not a surface filter.

Takes some work, but it is possible...

Nice work 3d..
I'll have to experiment with this when I get a chance...

So, here is how it works.

I started by rendering 4 different textures in FF. In this case, Ayeaka's Stone Flooring, Crapadilla's Rough Plaster, Kochubey's Parquet and Vlad's Decorative Tiling. I also rendered their alpha channels, although that is not visible from the end result.

Then I took each image, loaded its corresponding alpha channel as a selection (Parquet didn't have one, so I loaded the image itself as selection) and run it through a filter that split the RGB channels and the selection into a two-by-two greyscale image. For instance, Decorative tiling looked like this:



(That's R, G, B and alpha split apart.)

So, now I had four images similar to the above. Let's call these "preprocessed" images.

Next I made a new document of the same size and loaded one of the four preprocessed images as a selection. Then I ran it through a filter that composes the RGB image back together and places it in one of four positions on the two-by-two grid. As in:



I saved this image and loaded it back in, loaded the next preprocessed image as a selection and placed it into the second position on the grid:



And the same for the remaining two.

You can see the end result in the previous post.

---

Okay, so this is a pretty complicated example, but you could just preprocess one image and load it as a selection in another image and you would have two input images. The only limitation is that the second image is half the resolution.

I'll throw these multi-image-input filters in the queue if anyone is interested.

Quote
ronviers wrote: Maybe this process could be used on quadrants of the same image as a way to generate masks to eliminate macro patterns in seamless tiles. Noise could be used instead of hand painting. Having trouble wrapping my head around it right now.

Uhm... what's a "macro pattern?"

If you scroll down to the middle of this page you can see that they reduce the problem by using overlay maps.

Here is something that reduces them a little. It might be useful for the second pass. Now I am thinking it will not help to run it in each quadrant.

mP b.ffxml

While we wait for multi-image support in FF3 I thought I would give this thread a bump and add a bit to it....Just a quick filter and similar to ThreeDees concept......This is the test image...composited 4 images to make one image for processing in FF....Use this one for testing if you would like....

CFandM thanks for the 4-Image Setup filter, interesting and very simple solution to get beautiful results

And another example