How do grayscale values work in an RGB document?

Question

I am trying to understand how channels work. I created a blank 8bit RGB document in Photoshop and filled the background layer with 50% gray. But however, when I navigate to channels and click on R, G or the B channel, I can see that the channels are not 50% gray. They are somewhat a lighter shade of gray. Why is there difference in value of RGB vs individual channels?

As a vfx artist, I work with displacement maps which come packed with different information in R,G and B channels. Sometimes I need to extract individual channels from the file. So what I am trying to understand is that when I extract a particular channel, do I need to perform any correction to maintain the luminosity levels or a simple - 'copy channel and paste it into a new document, save as grayscale' is enough?

Answer 1

For historical reasons, VFX and print design have different conventions. Mainly because the way computer graphics was invented in roughly two different places at the same time.

So to the problem is that a displacement map, a depth map or a normal map do not really make sense from a the perspective of a software that is meant to make images for screen or print. Because they are just data containers, not images as such. You can off course work with them through channels palette, but they wont display correctly, because they are not images not really they cant be displayed correctly without a 3d rendering environment. This does not change the data one bit, just what it looks like.

So what is the problem? Color on your monitor is not linear, depth maps, displacement maps and normal maps are! This makes it a bit complicated for you to work in Photoshop. You can mitigate this as follows:

• You could obviously just tell photoshop that your gamma is 1 for all systems. This is the correct way to do this. But its probably going to shock you beyond belief. Because all your interpretation of color so far turn out to be wrong.

  The benefit of this is that now photoshop knows what your image is like so now all operations work correctly on a mathematical level.

• You could turn all of your transfer functions to 2.2 gamma. Or anything else for that matter.

  The downside is that you now have a hard time truly understaning what is going on and risk operations like blur working out wrong (although you can adjust this setting).

• you just ignore what you see and don't operate on them, the data is the data irrespective of how the what you see is what you get layer tries to convey otherwise. Its not an image it does not show right anyway no matter what you do.

  You can still work on moving data between places but not use photo shops operations without risking correctness.

But yeah next question i suppose is going to be what color does a transparent pixel have? According to Photoshop it does not have a color but in VFX it does, and its important that it does!

Answer 2

I will try to make some rudimentary explanations about color representations. But I can not answer your final question if one copy-paste for your workflow is enough.

Color, specific

Color is not absolute, it is relative, and in fact, it is very device-specific.

Take one cyan water marker and use it on magazine-coated paper and on a newspaper. The information about the color is the same (Cyan 100%), even though the ink is the same, the color is not.

Color visualizations

That is what color profiles (and color spaces) try to predict.

Normally, on print-oriented software, or at least oriented in the initial settings some visualization corrections are made. For example, a single channel gray only file is visualized as a lighter tone as a desaturated RGB file, because it is assuming some printed conditions, a single ink, instead of an RGB file that will throw several combined inks during the print.

Or the RGB colors could look less saturated and vivid because it is previewing the colors as they will render when converted to CMYK and printed.

---

Transformations

But that is only the beginning. The visualizations. Then we have the transformations.

These are different ways to how a color will best fit together with all the other colors inside a new color space or mode. As color is relative to other colors there are different ways to squeeze them.

Will this grayish-looking black, should be transformed as such, for when new conversions come, it never deceives the viewer to expect a pure black?

---

So, probably (I am more and more ignorant as this answer progresses) you don't need any transformation, but the raw numeric data. So try changing the default settings on grayscale images to gamma 1.

I assume that when you open an image you should not convert to the new color profile, but properly establish a defined working profile setting.

Answer 3

The RGB channels shown as separate are grayscale images and obey what's the selected grayscale gamma in the color settings. That can be useful for some print jobs but cause easily confusion and errors in RGB-only jobs. The RGB numbers get easily changed if you copy a single channel content from one to another sRGB image.

Stop the confusion. Change the grayscale gamma in color settings to custom value 2.2 That's the same as what's settled in the prehistoric days for RGB displays. After making that change the default lightening from 128 to 160 you have noticed, vanishes.

GIMP and probably also other RGB-oriented graphic programs do not have this problem.

Beware: One day you may need the original or at least some different setting than gamma=2.2 for predictable grayscale print works.