2

In Photoshop, it's fairly easy to separate an image into separate monochrome red, green, and blue images. These could each be projected as slides in front of separate red, green, and blue lights to form a color image.

Let's say I need to make monochrome slides like this for a projector that, for some reason, uses orange, teal, and purple lights. How can I separate an image into three monochrome images for orange, teal, and purple channels?


I'm working on a project about how color vision works, and part of the demonstration I'd like to be able to show involves what would happen if our eye's cone cells were sensitive to different colors.

  • Color theory is a bad tag as it does not really talk about how color i to be handled mathematically. – joojaa Aug 4 '17 at 8:22
  • Also note that this transform will not demonstrate how it would look like if we had different primaries, and computer graphics might be a better venue considering the audience. You could try to just do a matrix transform – joojaa Aug 4 '17 at 8:46
  • See: graficaobscura.com/matrix/index.html – joojaa Aug 4 '17 at 9:02
  • Ok this is definitely more suited to computer graphics.SE Because the site supports TEX and makes it nicer to read the resulting math. – joojaa Aug 4 '17 at 10:11
  • 1
    @Stan that is right Thats why in the last two paragraphs i am saying it will not actually work. And that one should in fact get a spectal image (hyperspectral or multispectral) image and integrate your hypothetical response curves if one wants to simulate a new type of creature. anyway this is well beyond graphic design. – joojaa Aug 5 '17 at 6:05
1

This is essentially what the color management engine does. What you do is you convert your color to linear, and then you formulate a the primaries as a matrix. You may want to formulate your color space using 4 properties [Red, Green, Blue and White point], or by 3 properties [Red, Green, Blue]

How it is done

The matrix formulation to 4 properties is described in length by Ryan Juckett although he describes how to convert a linearized sRGB to LAB conversion. It suffices to substitute lab primaries with your own. I am leaving this derivation for a different forum since it is mostly out of the scope of this site, Photoshop can do this (see below). Most graphic designers aren't so fluent with Linear algebra. Remember to convert to whatever non-linearity your lights expect.

The formulation with 3 properties is quite similar. The first step is to linearize (again, since linear algebra does not work with nonlinear values) except you just plunk your primaries into your matrix (as row or columns, your choice). invert it and use that matrix for conversion. Convert to nonlinear again.

Doing this in Photoshop

Now as I previously mentioned this is what color profiles do. Photoshop has a rudimentary way to define custom color profiles. Anything more complex you'd need to make the profile yourself as described above in the math section.

This is what you do:

  1. Choose Edit → Color Settings...
  2. In this dialog enable More Options, if it is enabled your fine and it reads Fewer Options
  3. In the Working Spaces drop down choose Custom RGB...
  4. A new dialog opens,

    enter image description here

    Image 1: Example of conversion dialog, insert your values.

    in this dialog write your:

    • RGB primaries and white point in device independent CIE XY(Z) Coordinates
    • Describe the non-linearity of your color space (Gamma)

    • Give this profile a name. Press OK.

    • Do a profile to profile conversion (Red is not your first primary and green second and so on)

Note this does not help you to see what things look like with different primaries!! (for one a lot of data was lost capturing and second your still looking with your primaries) You can only get channels that show intensities at these primaries. But if you want a better emulation, turn the channels to spot colors. But this emulates how humans see with human primaries

But this wont work for simulating different organs

Real environment data is not like the one you see in images. What we call color is only a sensation. If you want to simulate a different sensing then you need to scan the world differently than you do now.

For one color capture in cameras assume you have a sensing organ like a human. Now a different creature might have different primaries but also different distributions of those primaries in spectra. See red does not exist in reality, it exist in your brain, there are many different distributions of light that give you a red sensation. Some give you the same sensation event though they are different (called metamers).

enter image description here

Image 2: Human color response for each color sensor

Shifting the elements around and their distribution changes what goes where and how its sensed.

If you really want to emulate this what you need is a spectroscopic camera, or get a spectroscopic image and then integrate that data by your own response curves. This is much more accurate, although how to show or how to process things that is another thing.

You can find hyperspectral images here If you want to do more serious work on this.

Whew, thats it. Good luck

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.