To quote Kyle’s Mom from South Park, “What, what, WHAT?! Have you gone mad, man?! There is no color in a black and white image!” Well no, there isn’t, but there is a correct way that those colors should be converted and represented in a black and white image.
Here, let me show you what I mean. Here are color images of three automobiles:
Now, here are three different common methods used to convert it to a black and white or monotone image:
All three look very different. Which one is correct? Are any of them? How can we tell?
When we strip away the color, is there anything left in that shade of gray that would be present in the original tone? Yes—the luminosity.
Every color tone has a specific luminosity, and yes, since there are more possible tones in RGB than there are are shades of gray, there will be duplicates. Some different colors will have the same luminosity. But there is some method to this madness, so lets take a look.
Let’s take some pure tones and measure their luminosity. We will measure it with our color sampler in Photoshop, using Lab Mode, which separates the luminosity from the color channels.
Here we have Red, Green, Blue, Cyan, Magenta and Yellow, and the luminosity values for each ( 0 = black, 100 = white)
Now let’s take some common methods of converting an image to black and white, and see what the results look like.
Here is what happens if we just desaturate the image:
Well, that certainly doesn’t work. It makes all the tones look exactly the same—probably the reason that desaturating is the worst method for conversions. It makes a monotone image very…monoshade.
Channel Mixer (Red Channel)
Here, we will use a common method—the Channel Mixer, using the red channel:
Well, that’s not much more successful, since it changes the pure tones to pure black or white. If I would have chosen the other channels, it would just have made different blocks of colors appear black or white, depending on the color channel.
Now, what we could do is mix the color channels in different proportions to arrive at our desired luminosity, but that certainly would take quite a bit of trial and error, especially if we were doing a complex photograph containing millions of different tones, not just six.
In this example, I used a Gradient Map to do the conversion:
Ok, that looks a little better, and at least we are showing some difference in shade for the different tones. But it’s still not perfect. I’m actually a little surprised that the gradient map didn’t work better, and the truth is, I am not really sure why. It may be because the gradient maps are very linear, whereas relative luminosity is not. Maybe the color and light scientists out there can explain it better thanI.(I make a good photographer, but I am a horrible scientist.)
And it also changed depending on whether I used a smooth gradient or a 10-stop stepped gradient. So again, it was possible to fix, but it took a lot of trial and error. And again, much easier to do with six tones than with a complex photograph.
So, is there any method that does it right without all the fuss and muss? Well yes, there is. There’s the ‘convert to grayscale’ action. Converting your mode to grayscale retains the luminosity of every tone; it just removes the color.
“Well, why didn’t you say that in the first place?” Because we would have missed out on all the fun of learning something, and also it would have made my post about two lines long and people still would have asked why all this happens.
If we convert to grayscale (I usually convert back to RGB so that I have more tools available to me in Photoshop) here are our results on pure tones:
Here are the real-life examples of our automobiles. I included the color example right under it, so that you can see how the overall balance of the image (from light to dark) does not change, even though we have stripped out the color:
Now, I am certainly not saying that ‘convert to grayscale’ is the only method you should use for your black and white conversions. Your artistic vision may make one of the other methods the weapon of choice. I was a big fan of using gradient maps before, but after playing with this and seeing the actual results, I have started to use ‘convert to grayscale’ as my method of choice