Mysterious Photoshop Blending Modes

By Ctein

Blending modes are one of the more opaque (ahem) aspects of Photoshop. They show up when you're fading adjustments or filters, using layers, using a brush, or performing Image Calculations. Most of the time, blending gets left set to "normal," and you'll never notice it. It's when you decide to use it that the fun (I use that word advisedly) sets in.

Some of the blending modes, like "Lighten" or "Darken" are pretty obvious. Most are less so, but a little bit of playing shows, for example, that blending with "Luminosity" affects only lightness or darkness while "Color" affects only, ummm, color. That's a useful pair. For instance, doing heavy sharpening or edge enhancement can cause chromatic sparklies to appear in color images and exaggerate color noise. Doing the same operations with a Luminosity blend doesn't have those nasty side effects.

I still use Digital ROC for a lot of color correction and restoration work, but it almost always produces far too much contrast for my taste. I apply ROC to a duplicate layer in my photograph. Then I duplicate the ROC'ed layer, and set those two layers to Luminosity and Color blends. I use the fade sliders in each layer to control how much of ROC's changes get applied. Most of the time, I'll leave the Color-blended layer near 100%, but the Luminosity-blended layer will get dialed back to 40–50%. Much better results!

Then there are all those obscure blending modes. I mean, what the hell does "Hard Light" actually do?! I've come up with a visual aid that helps me figure out what's going on. It may or may not work for you.

Start with a 512 by 512 white canvas. Fill it with a gradient that runs from 0 to 255, left to right (Figure 1 upper left), or the full spectrum gradient (Figure 1 upper right) depending on whether you're more interested in black and white or color.

Figure 1

Now create a new layer and fill it with a gradient that runs from 0 to 255 from top to bottom (Figure 1, lower left). If you want to get fancy add a third layer and draw in some black pencil lines at the midpoints (Figure 1 lower right). Some blending modes use middle gray as a neutral/no-effect point, so the crosshairs make it easy to see where that level is.

 

Figure 2

The result will look like Figure 2, which is not very interesting, because what you're seeing is pure layer 1 at this point. What you have built, though, is a grid that superimposes every possible pair of values in black-and-white (and gives you a representative sampling in color). By messing with the blending, opacity, and fill controls for layer 1, you can explore the effect of different blending modes on different combinations of tones and colors.

Figure 3

Let's take a look at "Color Burn" to see how this works. Figure 3 shows what my black-and-white and color gradients look like when I apply Color Burn blending to Layer 1. How do you make sense of this? Each horizontal row of pixels in the grade is a gradient running from black to white (or through the spectrum). As you move up the grid, you see the effect of blending a particular pixel value into that, starting with 0 (pure black) at the bottom. Now the black-and-white grid starts to make a little more sense. When the blend pixel value is zero (bottom row), every pixel is set to black. When the blend pixel value is 64 (a quarter of the way up the grid), all the underlying pixels that have a value below 192 go to black; all the pixels above 192 (normally light gray to white) get remapped into a tonal range running from full black to white. As you move up the grid and the blend pixel gets lighter, the threshold moves further and further to the left, until at 255, the blend pixel has no effect.

Applying this to a color image (Figure 3 right) produces a strange look, but it's just the same thing going on in RGB channels. When the blend pixel values are very close to black, the color channel values get squeezed to 0 or 255. So, the rainbow values in the gradient get squeezed towards the primaries, where two of the RGB channels are at 0 and one is at 255. As you move up the grid, the compression effect falls off and the colors are less forced to red, green or blue.

Play with this for a while and see if it helps you wrap your head around blending modes. Many of them are a lot, lot weirder and less comprehensible than Color Burn.

Here's a bonus question worth extra points. If you think you understand what's going on in the grids with Color Burn, see if you can figure out what Color Dodge will do to the rainbow gradient before trying it. (I couldn't, but as soon as I looked at the results they made sense to me.)

Ctein

Comments

[vk]

Oh, no... not here...

[David Littlejohn]

I worked out all of this stuff for myself about four years ago, with a spreadsheet for the calculations and sample images. I'm sure you did a better job with it, but I was on the same track. Then, a couple of years ago, I switched to Lightroom and blissfully forgot all about this stuff. I don't miss it at all.

[Robert P]

I think one of us needs to go have a lie down.

[Stephen Best]

Many seem to have a fundamental misunderstanding of how blends work. A blend (in Luminosity or Color mode) takes that component of the layer and merges it with the layer below. For example, applying a Curve in Luminosity mode is not the same as applying that Curve to the image's luminosity. The former results in distortions as the luminosity relationships of the original's colours isn't preserved. Your Digital ROC approach suffers from the same fate.

If you're interested in exercising blends for your own use, check out the write-up I did on Lobster 2.0:

http://www.macquarieeditions.com.au/?p=88

[Aaron]

I'm sure I won't be the first to say this - that's really interesting to scroll through and see what the modes do, and try to figure out why they're doing it.

I'm not sure how I'll use that knowledge in my photo editing - that's another big leap from seeing what they do in your exercise to predicting what they'll do to a photo.

I have a few go-to blending modes now. If I'm fooling around, I fall into the "spin the mouse's wheel up and down" camp to decide what blending mode works best.

[The Worst of Perth]

The most opaque part of the blending modes is the obscure instructions in PS help.

[Bruce Robbins]

Usually, it's not really possible to predict accurately what effect you will get with different blending modes until you apply them so I don't think there's a lot of value in this sort of exercise beyond the academic.

For my own work, I find that I normally use just the first three "blocks" of modes beneath the normal blend. The first block of four darkens the image, the second block of four lightens it and the third block of seven increases contrast as you go from the top of the list to the bottom.

That's the way I remember it and it makes it easy enough then to experiment with a mode from the relevant block.

[ctein]

Dear Stephen,

I don't believe I asserted anywhere that using a Luminosity blend is the the equivalent of operating directly on a Luminosity channel. In fact, when I've written about Lab space, I've made sure to provide examples where Luminosity blends produce visibly different results from working in Lab space, so that readers understand they don't produce the same results.

It's not particularly germane to this column.

It's not germane to DIGITAL ROC, either. ROC only works on RGB images. There is no option for applying it directly to a Luminosity channel. Furthermore, preserving the luminosity/color relationships of the original image would be antithetical to ROC's purpose."Distortion" is meaningless in this situation.

The blending approach I use with ROC works extremely well. Before asserting it doesn't, I suggest you try it.

~ pax \ Ctein
[ Please excuse any word-salad. MacSpeech in training! ]
======================================
-- Ctein's Online Gallery http://ctein.com
-- Digital Restorations http://photo-repair.com
======================================

[John Roberts]

This article reminds me of the old man who used to say, "When I ask you for the correct time, don't tell me how to build a watch."

[Manish]

A very good explanation of blending modes is given in the book 'Grokking the Gimp'. See here:
http://gimp-savvy.com/BOOK/index.html?node55.html

[paul bailey]

Good stuff Ctein. I must confess, I get seriously dizzy when I venture into this depth of PS.

I'm only now doing "refine edge" instead of "feather".

frankly, I have found Capture NX2 a lot more intuitive, but you can go away for an hour and have a nap in between some actions.

[Chuck Kimmerle]

Interesting exercise, though I still cannot figure out what Pin Light does. I find the color artifacts produced by a couple of the blends a curious anomaly, especially when working with gray tones. Would these be caused by mathematical errors in the algorithms?

Stephen, is your assessment equally valid with adjustment layers as it is with image layers?

[Bahi]

It's useful to have your suggestions for images to use as a starting point: that might turn out to be the key for me. These esoteric-seeming blending modes have always looked vaguely promising but have been just that little bit too far from anything my intuition could help with, at least with the type of stuff I was trying with them. I'll start with your simpler image suggestions and see how I get on.

I was reminded of these out-of-reach (for me) blending modes again the other day when reinstalling Pixelmator, which uses Apple's Core Image to make these modes available under pretty much the same names as their Photoshop equivalents. I hope they do the same things but that'll be an experiment for another day.

[ctein]

Dear Chuck,

In the hopes of putting the lid back on a can of worms that I would rather not see opened further...

Luminosity blends in RGB space do not produce precisely the same results as operations performed directly on the luminosity channel in Lab space.

Usually the differences between the two are not of practical import. It can matter when the data is near the extremes of RGB (very light or dark or saturated).

If you get PHOTO Technique, there's an early 2006 article by me which includes some examples of this.

You should be aware that RGB versus Lab is often a religious war waged by partisans of the same ilk as PCs versus Macs, Nikons versus Canons, and Catholics versus Protestants. As in all such wars, the most forceful arguments are the least likely to be true, and both sides exaggerate their virtues, cherry-pick their examples, and belittle the opposition beyond justification.

I am not a player nor sympathizer in such wars and would rather not see them take over this column. Partisans should take their battles to the USENET newsgroups, thanks!

(and just to make sure I'm clear about this: I don't consider Stephen to be one of those fanatics, I am merely fearful he has unintentionally given them an opening....)

~ pax \ Ctein
[ Please excuse any word-salad. MacSpeech in training! ]
======================================
-- Ctein's Online Gallery http://ctein.com 
-- Digital Restorations http://photo-repair.com 
======================================

[chuck kimmerle]

Ctein,

Ummm....I think you read way too much into my last post. I was not fishing, therefore did even bring any worms.

Chuck

[Stephen Best]

Chuck:

"Stephen, is your assessment equally valid with adjustment layers as it is with image layers?"

Yes. Sharpening on a layer and using a Luminosity blend is not the same as sharpening the image's luminosity. With the former, you still get the smearing, it just takes the luminosity of the result.

Ctein:

"Luminosity blends in RGB space do not produce precisely the same results as operations performed directly on the luminosity channel in Lab space."

In Lab, it's Lightness. Different thing.

[ctein]

Dear Stephen,

Dammit. I *always* do that! I know it's "Lightness" and it always turns into "Luminosity."

Thanks for catching that.

Got a private email from Chuck. There's been a bit of a misunderstanding: his problem is something other than what we've been talking about-- he's seeing 'color artifacts' when he does some blends on crossed gray gradients.

I've rechecked all the blends, and I don't see any color anywhere (and wouldn't expect to). I think he's got a monitor calibration problem. Have you any thoughts on what might be afflicting him?

pax / Ctein