Iuvenis_Scriptor opened this issue on Jul 06, 2021 ยท 5 posts
Iuvenis_Scriptor posted Tue, 06 July 2021 at 2:25 AM
I'm about at my wit's end with this, so I decided to see if anyone here might have any ideas. I've been experimenting with storing what amounts to three trans maps into just one by applying a single pure color to the white areas in each one via Photoshop and then using a channel separation node in the Material room to generate the separate masks I need. The problem is that, it seems that no matter what I do, the result is always at least slightly different from the respective original black-and-white maps. Namely, there are these mysterious gray areas in each separated result, and as the subtraction in this example helps make clear, what should be pure black or pure white is often an off shade of gray. This suggests that there are tiny amounts of each primary color in areas that look either black or a completely opposite hue to the eye, but I cannot seem to track down the source of these spurious regions in Photoshop. I think the very way the maps are made should, in theory, prohibit such color bleeding, but I'm no expert.
I've tried various gamma settings, various methods of map creation, and even recombining the components and then re-separating the output. The one thing that seems to really reduce but not entirely eliminate the problem is using a 32-bit color format like EXR, but given that that's automatically a bigger file, it sort of defeats the purpose (i.e. minimizing storage space occupied by the maps).
Thanks in advance for any advice!
bagginsbill posted Tue, 06 July 2021 at 6:56 AM
When I sample your screen shot above, I found many places where there's red and blue mixed into the green, red and green mixed into the blue, etc.
You said "I think the very way the maps are made should, in theory, prohibit such color bleeding" but didn't communicate "the very way" so I'm going to guess that you have some sort of color space accidentally chosen that is introducing some hue shifting. Photoshop is not a good application to be using for "data" manipulation - it is far too prone to doing things "because reasons" that are not literally what you told it to do.
Renderosity forum reply notifications are wonky. If I read a follow-up in a thread, but I don't myself reply, then notifications no longer happen AT ALL on that thread. So if I seem to be ignoring a question, that's why. (Updated September 23, 2019)
Iuvenis_Scriptor posted Tue, 06 July 2021 at 10:52 AM
Thanks, BB! I've actually tried a couple of different waysways. I have what should be the white areas in each traditional trans map in separate layers on a black background in a PSD. I've tried applying color overlays to those layers. I've tried creating a black background and three pure hue layers and then applying the original trans maps as layer masks. Nothing seems to work.
As we speak, I'm using color math nodes in Poser to turn my three traditional trans maps into an RGB-channel-based combo and rendering it on a square prop. Let's see if that works. If not, do you have any alternative software that you'd recommend for this sort of data manipulation?
EDIT: Nope, that didn't work either.
bagginsbill posted Wed, 07 July 2021 at 3:26 PM
I should think Poser can do it just fine. Not sure why you were using color math. This is the easiest way:
Renderosity forum reply notifications are wonky. If I read a follow-up in a thread, but I don't myself reply, then notifications no longer happen AT ALL on that thread. So if I seem to be ignoring a question, that's why. (Updated September 23, 2019)
bagginsbill posted Wed, 07 July 2021 at 4:00 PM
I think that any of these Python image toolkits could easily combine images into an RGB as well. I believe PIL is already installed with Poser Python.
10 Python image manipulation tools
Renderosity forum reply notifications are wonky. If I read a follow-up in a thread, but I don't myself reply, then notifications no longer happen AT ALL on that thread. So if I seem to be ignoring a question, that's why. (Updated September 23, 2019)