Depth influenced composing

[kirkl]

I like some Photoline features and would choose Photoline over Photoshop probably once I would be able to do a complex, depth combining image composing like this one:
https://dl.dropboxusercontent.com/u/757 ... btract.psd




Set "Move tool V" to layers instead of groups and move green star.

PL needs a to either have Photoshop styled transform links in-between layers or a kind of layer selection sets or maybe some formula based layer calculations. Do it more elegant way than Photoshop as everything else please.

[Hoogo]

I guess something got lost during import of this PSD to PL? At least I don't get the idea when I look at the layers.
Tricky one, I have some ideas, but are not completely there yet

What I already have:
-Combining 2 depth maps into a mask: Subtract + adjustment layer "Threshold".
Moving works as long as you select the green graphic and the green depth. That's what the attached ..3.PLD can do. I had to use Jpg compression to keep the limit of 128K, but it looks good enough.

If you move often or have a lot of this kind of object, then it would be more practical to move only one layer.

What I know how to do:
-If you create a virtual copy and move the original, the copy stays at it place.
-I you group before that and create the virtual copy from the group, you can move the content, and the copy moves, too. But that didn't help here.
...but that way you move only one picture, not 2 different ones.

It would work if graphic + depth were coded into one "layer":
-If the objects have such simple colors like here, then you could store graphics and depth in channels. The channel mixer can select the channels where needed.
-In 16Bit you could use parts of the tonal range to store multiple 8Bit pictures. The histogram correction can reveal and select the right one.
-I guess the best solution: Place picture and depth side by side and use a mask to hide the unwanted one.

I you think that these solutions are less crazy than the PS version, I'll try an example with the last suggestion.

PS: Is this something like a common topic, pictures with depth? I suggested a Z channel when the 3D hype started a few years ago.

[Hoogo]

Just checked your other topics.
I did not understand them all,but I guess the side-by-side thing could be a workaround for many of the topics that need more channels.

[kirkl]

Yeah, a dedicated depth channel for each layer would be super cool and easy. Corel Painter has one for example, a true 32bit depth channel they use for impasto brushes. But for some weird reason you can't do anything useful with it , no depth combine, nothing to even combine those impasto strokes properly on different layers. And you can only export it from Painter files through some Blender addon.

Still why can't we do it regular way. The only thing to having it working more or less comfortably is a Photoshop styled "chain" links in between color and depth layers. In property panel maybe. So I could select and move only top visible group without looking where is the second "depth " part each time.

And in fact I also need one more "roughness " channel at least. Since all this is for composing modern materials/textures and background planes for 3d scenes. So such "chain" links would be very helpful in other cases too. In fact I find Photoline masking stack a lot more elegant and logical than Photoshop groups clipping I used in my psd example. PL really misses those "chains" only.

ps. I tried the idea of packing the depth into the range of one of rgb channels but broke my mind already to do it properly. If you have a working example I would be appreciative to see. Perhaps should try it with 32bit and store the depth somewhere beyond 0-1 range? Although would be working too slowly probably.

[Hoogo]

ps. I tried the idea of packing the depth into the range of one of rgb channels but broke my mind already to do it properly. If you have a working example I would be appreciative to see.

I tried it with reducing 8Bits into 4Bits and combining them into the blue channel. I used gradients instead of the original pictures, that makes testing the results much more easy.
It already comes close, but either I'm often off by one, or I have some fractions left. Have to try layer mode "divide", but I already have the feeling that it's not worth it. The idea isn't that difficult, but expressing the formula by layer is.

[kirkl]

Thank you very much Hogo. I think your approach could really work, especially with 32-bit. But looks like a bit too complex doing a lot of otherwise redundant range splitting operations. Would be slow, I bet , with real images.