I’ve got a bunch of rendering to do that involves a small sport clothing shop with lots of apparel and products all over the place. My client is very picky about the appearance of the product colors. Down to matching particular PMS colors.
My problem is getting my renders to match the product colors as closely as possible. In the past I’ve just used Photoshop to do color corrections on specific colors. But for this project I’ll have to do multiple variations on fixture designs with about 6 cameras in the shop, for a total of about 24 different camera/fixture configurations. If I were to create alpha channels for each color that had to be corrected for each shot I’d working on this project for the rest of the year.
I find that under Color Mapping, Linear Multiply produces the best color match but requires much more fiddling with the lighting setup to prevent blow-out and to get a good distribution of light.
With the other color mapping options - Exponential, HSV Exp, etc - I find that color tends to wash-out and really shift around the spectrum. For example there are some orangy-yellow tee shirts that render out more of a green-yellow. I’ve tried using a correction layer in Photoshop to globally adjust the color value, but I can’t seem to find the right combination to affect all the colors correctly. I need to adjust each color separatley. Of course the thing I like about Exponential is that I seem to get a better light distribution. Not sure if that’s the right term.
So I’m wondering if there is a way to get better color accuracy using Exponential color mapping. Either with settings in VRay or with some sort of correction in Photoshop. Or should I just stick with Linear Multiply?
I wish I could post an image but I’m under one of those dreaded NDA’s.
Here is what I would do in your case. Try making the best lighting with the best colors.
Then switch to Scanline render, and make pure R,G,B renderings on the objects you would need to adjust manually in PS. Remember to delete all the lights and use pure self-illuminated materials for masking.
That works quite fast for retouching a render, as this gives you power control selecting the R,G,B channels only. Rendering the masks would take seconds.
Its allways the same with clients dont you think so ?
there´s no way to get any real color
an immage is allways the interaction of light in lots of colors with surfaces.
or to say more clearly, what we call a color is the permanent behaviour of light “interaction” of a surface.
to understand a color , our brain recalculates allways the whole scene with all the interacting light, to extract this reflection behavior of surfaces.
so a white paper on its own can be bluish or more red, but never has an exact rgb vallue.
you have to make clear to your cllient that this allways is a misunderstanding of tru behaviour. of colors and light.
also make clear to him, thaat rendering a photorealistic and also a hyperrealistic immage is not the reality, cause you have allways to deal with the same problems as a photographer..
so you open your shutter of the camera a little bit longer and everything is brighter… now the question, which photo shows the real brightnes and the real color ?
of cours the answer is none of them ore both. its just a photo…
but sure.. most clients dont mind about theese details, and won´t listen.
I agree with Tom, however I have found that I can match color almost exactly since using the linear workflow. At least I can say this: Whatever color I put in the diffuse (using color correct) comes out very believably in my renders without any PS adjustments.
Nikki’s idea would help in masking, but if you’re gonna go that way, you might want to look into PSD manager.
I also agree with tom, and just want to add that there is a lot of psychology involved.
The funny thing is that we perceive the bluish or yellow paper still as white - i dont know the correct translation for it, but it is called color constancy (farbkonstanz) in german
Hence it is more or less impossible to show “correct” colors in your pictures as the real object’s colors will vary anyway.
If the shop you have to visualize already exists you could try to measure the major lighting (shoot some HDRIs or place white/grey styrofoam balls in the scene for reference) - your rendered colors will more or less match their real counterparts in this case.
Otherwise (i.e. you have to visualize virtual lighting setups) you could base your lighting on real light properties (IES lights for example) to show what effect different lighting would have on the colors.
btw: where do you get the colors from? for maximum reliability the values should be recorded under normed lighting conditions (D65)
We are in the same situation here, all the time, so we are doing it a lazy and efficient way.
We are doing packages of 6 cameras with 10 corrections each, so 60 camera/color configuration, and no, it does not take an year, the total process takes about 2 hours, including masks generation (max), correction (Pshop) and print.
Our workflow:
(we can not use mat ID or other standard stuff, it is not precise enough, our prints are A3 format on a photo machine, not a printer, so we can not have that baddly antialiased pixel along the way)
Save as (we will be going back to good’old scanline)
right click > Vray scene converter > yes
MAXscript > run script > load convert_to_masks
right click > black masks > yes
What we do have now is all the materials standard, black, 100%self illuminated
Delete all lights in scene
Select the object(s) you want to change color
Set is material to white 100% self illuminated
render with scanline
You now have a crisp, perfect, selection mask
Open Pshop
Your original image, and select the new black and white image as your first alpha channel of correction
It is a pain in the a** to do over and over, but that’s the most efficient way we found yet for our use