Luminosity and Blend Modes

[KevinStecyk]

Does anyone have a good grasp of how Luminosity and Color Blend modes work in Photoshop CS3 (same as CS2) and can explain it in layman terms? Note: Title of this thread is missing the word "Color".

I am having difficulty obtaining a good understanding. If possible, I'd like an explanation in rather simple terms. I am getting the impression that the true answer is quite complex.

To demonstrate my difficulties, please do the following:

1) Take a colorful photograph (any colorful photograph will do)

2) Add a new layer with a large rectangle covering most, but not all of the picture below. Color this new rectangle some color (mine is yellowish with R241, G236, B71)

3) Add to this yellowish layer 5 colored stripes (RGB):
3a) Black: 0,0,0
3b) White, 255,255,255
3c) Red: 255,0,0
3d) Green: 0,255,0
3e) Blue: 0,0,255

4) Look at your image with the top layer at different blend modes:
4a) Normal
4b) Luminosity
4c) Color

Luminosity:
-top layer provides detail
-bottom layer provides color

Color:
-top layer provides color
-bottom layer provides detail

Yet, look at Black and White stripes. In Luminosity mode, the underlying color is blocked. In Color mode, these stripes appear nearly transparent.

Can someone provide an easy-to-understand explanation of what is happening in these different blend modes?

[KevinStecyk]

I got some help from others in a different technical forum. I thought I would share my learnings.

Stephen Marsh wrote:

======>>
Here are the RGB Luminosity (gray) values for eight key test colours
(Adobe RGB) -

White: 255 (0%)
Black: 0 (100%)
255 R: 76 (70%)
255 G: 150 (41%)
255 B: 28 (89%)
RGB C: 179 (30%)
RGB M: 105 (59%)
RGB Y: 227 (11%)

Notice those last three idealised CMY % values? 30 59 11 ...

Sounds a bit like 30r 59g 11b, which is close to Photoshop RGB Luminosity weighting (not precise though, just close). Strange how that worked out, very strange indeed! Coincidence, or something deeper? Anyway, enough of that!
<<=======


I then wrote:

=======>>
As you suggested (and what I should have done initially), I looked at CS3's help file.

>>Luminosity: Creates a result color with the hue and saturation of the base color and the luminance of the blend color. This mode creates the inverse effect of Color mode.<<

...

So looking at Black and White, the luminance values are 0 and 255. In RGB, luminance of 0 is always black. So the underlying color does not matter. Similarly, luminance of 255 is always white, so the underlying color does not matter. Am I correct?

For other colors in between black and white, their luminance values will be between 0 and 255. Thus, the underlying color does show through and the underlying colors do appear to affect the luminosity. For example, Red in sRGB has a luminosity value of 54. Yet, when I use my file (mentioned at the top of my response) with Luminosity blend mode, the luminosity varies from the low 30s to high 30s.
<<=======




Dan Margulis then wrote

=======>>
The above summary is correct. It is not possible to have an RGB color (meaning, other than white) that is as light as 255r255g255b, because to introduce the color one or more of the channels would necessarily darken. Similarly there cannot be color that is as dark as 0r0g0b because adding the color would necessarily lighten one or more channels.

In LAB such colors can be made theoretically and there are several uses in retouching for them. If you convert the files described above to LAB without flattening, you will create such imaginary colors. Photoshop displays them as being neither very colorful nor absolutely light or dark, but rather a combination of both. If you then flatten and reconvert to RGB the LAB look will hold.

> For other colors in between black and white, their luminance values
> will be between 0 and 255. Thus, the underlying color does show
> through and the underlying colors do appear to affect the
> luminosity. For example, Red in sRGB has a luminosity value of 54.
> Yet, when I use my file (mentioned at the top of my response) with
> Luminosity blend mode, the luminosity varies from the low 30s to
> high 30s.

This is just an extension of the white/black situation. When what's being called for is near the edge of what's achievable then it is likely not possible to retain both color and luminosity, so there will be some compromise.

255r0g0b is an example. This is the reddest red that RGB can produce, and it occurs only at one specific luminosity. If the file becomes either lighter or darker some of the redness must be lost.

Consequently, Luminosity/Color modes whether in RGB or LAB are only truly reliable with subtle colors. They do not work well when painting with near-whites or near-blacks. When brighter colors, or near-whites or near-blacks are in play, RGB and LAB behave differently and either (or neither) may be the result that's wanted.
<<=======


Dan Margulis is the author of the following two books:

1) Professional Photoshop: The Classic Guide to Color Correction (5th Edition)

and

2) Photoshop LAB Color: The Canyon Conundrum and Other Adventures in the Most Powerful Colorspace

Note, the two above links are Amazon sponsored links.


I am posting this information so that a) others don't spend time and effort addressing my question when it has already been answered, and b) to help others who might be interested.

[Noel_Carboni]

Imagine an image with two layers. The one at the bottom is your work in progress, and the one above is a derivation of something you've done, and you've set the mixing mode of the upper layer as follows:

1. Luminosity.

For each pixel in the upper layer, the luminosity is calculated, and combined with the color from the lower layer.

This can be useful for sharpening, for example. You can get color artifacts if you sharpen an image heavily, but if you blend the sharper data with the original color, those artifacts are averted.

2. Color.

Just the opposite from the above: The luminosity from the lower layer is combined with the color from the upper layer.

I find this useful when adjusting color saturation results in an unwanted change in luminosity. Many times, for example, when working with blues and cyans changing the saturation can yield unwanted artifacts. Changing just the color, but using the original luminance can be very helpful here.

Now let me tell you about a good trick to employ the above:

Run your filter or whatever on your image in just one layer, then choose Edit - Fade, then pick either Color or Luminosity. This has the effect of applying the filter only to the color or luminosity of the image. Conceptually the same as splitting the image into layers, using the filter, choosing one of these blending modes, and recombining the layers, but much faster, and you stay in RGB mode.

Make a habit of choosing Edit - Fade - Luminosity right after sharpening, for example, and your sharpening quality will likely improve.

-Noel

[KevinStecyk]

Hi Noel,

Allow me to frame my demonstration better.

Here is a photograph of Lake Louise, which you've seen before, as my background layer.

If you click on the image, you will see a larger image.



Now let's look at the different blend modes. I have added a layer with a color of R241, G236, B71. Then I added the following stripes:

1) Black,
2) White,
3) Red,
4) Green,
5) Blue, and
6) Gray (r128,g128,b128).


First we have the normal blend mode.

Originally Posted by Noel_Carboni

1. Luminosity.

For each pixel in the upper layer, the luminosity is calculated, and combined with the color from the lower layer.

Not exactly. Here's the luminosity blend mode.



Look at the Black and White Stripes. It doesn't seem to me that the color from bottom layer is being picked up. Also, note how similar the Green and Gray stripes are.

Originally Posted by Noel_Carboni

2. Color.

Just the opposite from the above: The luminosity from the lower layer is combined with the color from the upper layer.

Again, not exactly. Here's the color blend mode.



Look at the Black and White and Gray stripes on the upper layer. They appear almost transparent. So we can conclude that the upper layer's colors do not always dominate.

This is the concept that I was struggling with.

Best regards,
Kevin

[Noel_Carboni]

Kevin, your examples perfectly illustrate the concept.

In the first example a black or white stripe isn't going to show any color because the luminosity is 0% or 100%, and there's no "room" for color. And the color is quite clearly showing through the other stripes. I think perhaps what you're having trouble grasping is that there are limits to what colors and brightnesses can be represented - this is called the gamut of the color space. If you're at the limit of luminosity - 100% - then you can't expand the color space to mix in a color. The result will stay white - 100% luminosity.

Likewise, please point out where the color isn't as expected in the second mix. The yellow background is coloring everything it is in front of various brightnesses of yellow, etc. The color of black or white is neutral, so the color of those stripes will be shades of gray. If you're not seeing gray then your eyes are being fooled; sample the colors and see if the R, G, and B values are not all equal.

-Noel

[KevinStecyk]

Agreed Noel.