Professonional Photoshop, Chapter 8: Keeping the Black and White in Color
mwgrice
Registered Users Posts: 383 Major grins
In this chapter, Margulis shows how to use channel blending to improve contrast in color images. As the title suggests, the techniques used in this chapter build on those used in Chapter 7 to make good black and white images.
The basic premise of the chapter is quite simple. "If you return here after finishing the entire book, " Dan says, " you will have to agree that this chapter is the easiest of the 20. You open the image, look for a channel that has better contrast than the other two, and plug it into a luminosity layer. Game over."
A luminosity layer is a a layer whose blending mode is set to luminosity. This blending mode tells Photoshop to use the luminosity of the top layer and the color of the bottom layer.
Let’s look at an example similar to the one Dan uses. The following image, shot on the other side of the bridge from Dan’s figure 8.1, contains one channel with a photo of the Arno in Florence and three copies of a red square with a pair of smaller green, blue and white blocks inside it. Each square is set to a different blending mode (normal, luminosity or color), which is written in black inside the square.
As you'd expect, the block in normal blend mode blocks everything below it. The block in luminosity blend mode has a high overall luminosity and often overwhelms the underlying scenery. However, where the luminosity of the block hasn't overwhelmed it you can still see the colors of the trees and buildings underneath it. Finally, the block in color mode has transferred its colors to the image. The writing and the smaller blocks are hard to see in the sky, where the luminosity from the lower layer is so much less.
(By the way, Rutt wrote a much better tutorial on luminosity and color blending modes here.)
What good does this do us? In practical terms, you can use any channel in an image with better contrast to improve the overall contrast of the image without having to worry too much about the color of the image. You can also use any copy of the image, such as a black and white version or a high contrast version with poor colors, to improve the contrast of the image.
Let me show you an example. A portrait will often benefit from a luminosity blend. Dan says that the best channel from a portrait is usually the green channel. Here is a portrait I took over Christmas:
Let's look at the individual channels:
Of the three, the red and the blue are clearly inferior. The red lacks details, while the blue is clearly overdone. So I'll use the green channel as a luminosity blend.
Making the blend itself is pretty simple. Create a duplicate layer and set the blending mode to luminosity. You can set the blending mode afterwards as well, but I find setting it first allows you to preview the effects of the blend. Additionally, you could perform the blend without creating a duplicate layer, but using a duplicate layer is much more flexible.
Now click Image, then Apply Image. In the Apply Image dialog box, choose the background layer and the green channel. Although the default is to use the merged layers as the source, Dan recommends always using the background layer instead.
Now her face is darker and has more depth to it. It's subtle, but it's there.
Let's look at those side by side:
I should stress that at this point I haven't attempted any color correction. However, I have made a better starting point for future color correction. The image still needs to have dark and light points set, and it needs curves to improve the skin tone.
Dan does not recommend using this strategy in any other color space than RGB. Doing luminosity blends from within CMYK doesn't work well because of the ink limit, and LAB is so different from RGB and CMYK. You can, however, certainly use channels from other color spaces.
Additionally, watch out for noise in the final result. The blue channel tends to have more noise than other channels, although you should look for noise in any channel before using it for a blend. In some circumstances, a luminosity blend will accentuate the noise to unacceptable levels, even if all channels in the image were fine prior to the blend.
Dan uses a couple of different strategies to avoid unwanted color changes with luminosity blends. Even though the image gets its color information from the bottom layer, changing the luminosity can and will change colors. Sometimes this is harmless and sometimes this is not.
The first technique is to change the type of blend used when applying the image. Instead of using normal mode, you can use darken or lighten mode. Darken mode blends only if the pixel in the top layer is darker than the same pixel in the lower layer, while lighten mode blends only if the pixel in the top layer is lighter.
Dan points out that skies are usually the best in the red channel, so doing a luminosity blend with the red channel will often make the sky more dramatic. However, this can cause problems for any red in the image. For instance, the sky could stand improvement in this image:
However, applying the red channel results in this:
The sky is more dramatic, but the red in the Peroni ad is completely washed out.
This can be fixed by using darken mode. When applying the image to the duplicate channel, simply change the blending mode to Darken:
The result looks like this:
Now the sky is dramatically improved, while the red in the Peroni ad remains about the same. By the way, this is why I usually set the blending mode of the duplicate layer before applying the image. By doing so, I can preview the effect of the luminosity blend before I click OK, and I can experiment with darken or lighten mode if the image calls for it.
Incidentally, this is one of those situations where a luminosity blend can cause noise problems. While the noise was present in the red channel before the blend, we've just made it much more noticable. Below is a 100% crop of the sky in the upper left corner of the image.
The solution for this is to flatten the image, create a duplicate of the new background layer and run the surface blur filter on the red channel. Move the image into LAB without flattening the layers. Now use the blend-if sliders to exclude everything that isn’t very negative in the B channel. Dan points out that few things besides the sky are. If any areas in the image actually are highly negative in the B channel, you can usually use a mask to exclude them without too much trouble.
Finally, in some images dramatic color changes from the luminosity blend can't be avoided with darken or lighten mode. For instance, if the subject of a portrait is wearing red or a similar color, a luminosity blend of the green channel will often darken the red significantly and undesirably. Applying the blend in darken mode won't help since darken mode will allow the red to become more dark anyway. Applying the blend in lighten mode won't help either, since lighten mode will stop the blend from improving the subject's face by preventing it from getting darker. This may be an issue in many other images containing red
In the following image, the painted face is a faded red. (I had a hard time finding a portrait I liked where the subject was wearing red.)
Looking at the channels, the blue channel has more detail in the rust at the bottom of the image. I think this is an important part of the image, and I would like to enhance it.
However, a luminosity blend darkens the red in the graffiti face excessively, so much so that it almost looks like new paint.
In order to use the blend-if sliders properly, you need to find a channel where the layer above is as different as possible from the layer below. Dan recommends turning the blending mode of the top layer temporarily back to normal mode while you fiddle with the blend-if sliders. This turns the top layer into a grey-scale image. As you move the blend-if sliders, any area excluded from the blend will show up in color again.
In this particular image, the green of the rusty areas at the bottom is significantly lower than the green of the graffiti face. So I can use this to exclude the red in the graffiti face from the blend:
In the final product, the color of the graffiti remains faded but the detail of the rust at the bottom of the picture is more pronounced. Additionally, the noise has not been thrown out of whack from using the blue channel.
You’re not limited to simply using the RGB channels. Any channel or copy of the image with the detail you want is a good candidate. Other possibilities include:
The basic premise of the chapter is quite simple. "If you return here after finishing the entire book, " Dan says, " you will have to agree that this chapter is the easiest of the 20. You open the image, look for a channel that has better contrast than the other two, and plug it into a luminosity layer. Game over."
A luminosity layer is a a layer whose blending mode is set to luminosity. This blending mode tells Photoshop to use the luminosity of the top layer and the color of the bottom layer.
Let’s look at an example similar to the one Dan uses. The following image, shot on the other side of the bridge from Dan’s figure 8.1, contains one channel with a photo of the Arno in Florence and three copies of a red square with a pair of smaller green, blue and white blocks inside it. Each square is set to a different blending mode (normal, luminosity or color), which is written in black inside the square.
As you'd expect, the block in normal blend mode blocks everything below it. The block in luminosity blend mode has a high overall luminosity and often overwhelms the underlying scenery. However, where the luminosity of the block hasn't overwhelmed it you can still see the colors of the trees and buildings underneath it. Finally, the block in color mode has transferred its colors to the image. The writing and the smaller blocks are hard to see in the sky, where the luminosity from the lower layer is so much less.
(By the way, Rutt wrote a much better tutorial on luminosity and color blending modes here.)
What good does this do us? In practical terms, you can use any channel in an image with better contrast to improve the overall contrast of the image without having to worry too much about the color of the image. You can also use any copy of the image, such as a black and white version or a high contrast version with poor colors, to improve the contrast of the image.
Let me show you an example. A portrait will often benefit from a luminosity blend. Dan says that the best channel from a portrait is usually the green channel. Here is a portrait I took over Christmas:
Let's look at the individual channels:
Of the three, the red and the blue are clearly inferior. The red lacks details, while the blue is clearly overdone. So I'll use the green channel as a luminosity blend.
Making the blend itself is pretty simple. Create a duplicate layer and set the blending mode to luminosity. You can set the blending mode afterwards as well, but I find setting it first allows you to preview the effects of the blend. Additionally, you could perform the blend without creating a duplicate layer, but using a duplicate layer is much more flexible.
Now click Image, then Apply Image. In the Apply Image dialog box, choose the background layer and the green channel. Although the default is to use the merged layers as the source, Dan recommends always using the background layer instead.
Now her face is darker and has more depth to it. It's subtle, but it's there.
Let's look at those side by side:
I should stress that at this point I haven't attempted any color correction. However, I have made a better starting point for future color correction. The image still needs to have dark and light points set, and it needs curves to improve the skin tone.
Dan does not recommend using this strategy in any other color space than RGB. Doing luminosity blends from within CMYK doesn't work well because of the ink limit, and LAB is so different from RGB and CMYK. You can, however, certainly use channels from other color spaces.
Additionally, watch out for noise in the final result. The blue channel tends to have more noise than other channels, although you should look for noise in any channel before using it for a blend. In some circumstances, a luminosity blend will accentuate the noise to unacceptable levels, even if all channels in the image were fine prior to the blend.
Dan uses a couple of different strategies to avoid unwanted color changes with luminosity blends. Even though the image gets its color information from the bottom layer, changing the luminosity can and will change colors. Sometimes this is harmless and sometimes this is not.
The first technique is to change the type of blend used when applying the image. Instead of using normal mode, you can use darken or lighten mode. Darken mode blends only if the pixel in the top layer is darker than the same pixel in the lower layer, while lighten mode blends only if the pixel in the top layer is lighter.
Dan points out that skies are usually the best in the red channel, so doing a luminosity blend with the red channel will often make the sky more dramatic. However, this can cause problems for any red in the image. For instance, the sky could stand improvement in this image:
However, applying the red channel results in this:
The sky is more dramatic, but the red in the Peroni ad is completely washed out.
This can be fixed by using darken mode. When applying the image to the duplicate channel, simply change the blending mode to Darken:
The result looks like this:
Now the sky is dramatically improved, while the red in the Peroni ad remains about the same. By the way, this is why I usually set the blending mode of the duplicate layer before applying the image. By doing so, I can preview the effect of the luminosity blend before I click OK, and I can experiment with darken or lighten mode if the image calls for it.
Incidentally, this is one of those situations where a luminosity blend can cause noise problems. While the noise was present in the red channel before the blend, we've just made it much more noticable. Below is a 100% crop of the sky in the upper left corner of the image.
The solution for this is to flatten the image, create a duplicate of the new background layer and run the surface blur filter on the red channel. Move the image into LAB without flattening the layers. Now use the blend-if sliders to exclude everything that isn’t very negative in the B channel. Dan points out that few things besides the sky are. If any areas in the image actually are highly negative in the B channel, you can usually use a mask to exclude them without too much trouble.
Finally, in some images dramatic color changes from the luminosity blend can't be avoided with darken or lighten mode. For instance, if the subject of a portrait is wearing red or a similar color, a luminosity blend of the green channel will often darken the red significantly and undesirably. Applying the blend in darken mode won't help since darken mode will allow the red to become more dark anyway. Applying the blend in lighten mode won't help either, since lighten mode will stop the blend from improving the subject's face by preventing it from getting darker. This may be an issue in many other images containing red
In the following image, the painted face is a faded red. (I had a hard time finding a portrait I liked where the subject was wearing red.)
Looking at the channels, the blue channel has more detail in the rust at the bottom of the image. I think this is an important part of the image, and I would like to enhance it.
However, a luminosity blend darkens the red in the graffiti face excessively, so much so that it almost looks like new paint.
In order to use the blend-if sliders properly, you need to find a channel where the layer above is as different as possible from the layer below. Dan recommends turning the blending mode of the top layer temporarily back to normal mode while you fiddle with the blend-if sliders. This turns the top layer into a grey-scale image. As you move the blend-if sliders, any area excluded from the blend will show up in color again.
In this particular image, the green of the rusty areas at the bottom is significantly lower than the green of the graffiti face. So I can use this to exclude the red in the graffiti face from the blend:
In the final product, the color of the graffiti remains faded but the detail of the rust at the bottom of the picture is more pronounced. Additionally, the noise has not been thrown out of whack from using the blue channel.
You’re not limited to simply using the RGB channels. Any channel or copy of the image with the detail you want is a good candidate. Other possibilities include:
- The K channel. Dan uses this channel in Multiply Mode in the image of Machiavelli’s tombstone at the end of Chapter 7, as well as in one of the images at the end of Chapter 8. Don't forget to look at the C, M and Y channels.
- A good B&W conversion. Simply open another copy of the original, make the B&W conversion, paste it on top of the color image, and then set the blending mode to luminosity.
- Curves designed to increase contrast as described in Chapter 2 without regard to color changes. This can be an adjustment layer, a duplicate layer or another copy of the image pasted on top, but either way it should be set to Luminosity mode.
0
Comments
Taken together, chapters 7 and 8 teach something very important: use channel blending early in your workflow to improve contrast and give yourself a much better starting point for what follows. In Dan's advanced class, he said, "So why are my images beating your images most of the time? That shouldn't happen. You guys are supposed to be advanced." That was the start of a 3 hour lecture on channel blending, basically covering the material from these two chapters. These days, I'd rather start by teaching this stuff than how to use RGB curves to improve contrast (as in Ch. 2).
On thing I learned from Dan's last class which I don't think is in the book: you can sometimes repeat these blends with good effect. For portraits, make a green luminosity blend as shown above. Then flatten. Then do it again. It surprised me to find out I got more of the same this way. I found this especially important red channel blends and skies, but I also used for the ballet shots from the fall where one dancer had very light complexion.
"Everything has its beauty, but not everyone sees it." - Confucius
I know there's a lot out there to learn with channel blending. I love doing red channel blends for enhancing skies. I love using channels as the basis for masks. I'm starting to sometimes use blends to fix tonal problems rather than shadow/highlights.
But, in the spirit of constructive disagreement, I've tried the green channel luminosity blend a lot of times for portraits and just never found that it really improved things much or was the most effective way to get my desired result.
First, there are often lots of colors elsewhere in the photo who's luminosity gets just totally thrown off with a green luminosity blend (greens will go way too bright and reds or blues will go really dark). It's often hard to fix this with blend-if when in RGB, so then you're forced to either take the green channel to LAB where blend-if works better or you have to resort to other means of limiting the effect.
Second, lots of times when I do the green blend, the image just doesn't change much and even the little bit of change I see doesn't really improve things.
Third, you can reduce the effect of a green channel blend with layer opacity, but you can't easily raise it's strength. So, if it doesn't do enough, it's not easy to make it stronger.
Fourth, I find it much, much easier to add contrast to faces with curves in a luminosity blend mode. It's quicker and more controllable. You have the strength of the curve, the shape of the curve, the amount of movement in different color channels and the opacity of the layer all as useful variables. If the effect I want on the face needs to be isolated from the background, I'll do a 15 second soft mask of the facial area to isolate the effect from the background or the rest of the subject. I generally try to do global changes before resorting to masking, but curves are just so much more flexible for contrast changes.
So, combining all these experiences, the green channel blend for portraits has just not become a useful part of my toolbox.
Homepage • Popular
JFriend's javascript customizations • Secrets for getting fast answers on Dgrin
Always include a link to your site when posting a question
That's interesting. I use it nearly every time. All the ballet shots get it. To each his own, I guess.
Here's a technique to create a more dramatic effect in the highlight/shadow balance for faces. Curves never seem to adjust the shape in quite the same way. It's a little complicated, but most of the setup can be put in an action and then layer opacity sliders adjusted to taste.
I use all three RGB channels: Blue in luminosity mode, Green in soft light and Red in overlay. Usually blue is about half the opacity of green, and red is brought up to whatever level emphasizes highlights without blowing out the image. Often this will desaturate the colors so I bring the original image back on top in color mode, then run Shadow/Highlights. You can push it for really extreme effects, or keep it subtle. Here's the portrait from MWGRICE's summary (great job, by the way!) with the layer set-up as displayed.
That's a black plate mask on the Shadow/Highlight layer, pulled from a duped image converted to CMYK. Sometimes the eyes get a little dark and I'll dab white onto them in a softlight layer, which is the case here.
The action sets up the first four layers and defaults to Blue at 30%, Green at 60% and Red at 0%. The blue value might need to be darkened or lightened, depending on the image; likewise the green. In this case I left them as is. The red plate is always lighter so an overlay blend will bring back highlights but in a way that curves can't accomplish, given that the two previous layers have darkened everything. The color layer may or may not be needed. There's a lot of flexibility in the way each layer contributes to the end result.
Then I copy everything to a new layer and run Shadow/Highlights. Once the action's set up, it's usually about a thirty second investment of time for fairly decent results.
—Korzybski
I have to say that going into LAB for the blend-if sliders doesn't bother me, since I'll be going into there for HIRALOAM. I may wind up going more than once, which doesn't bother me.
I think Rutt partially addressed your concern about strengthening it--flatten the image and repeat. The obvious downside to that would be color shifts from luminosity changes, since I think you'd have to repeat the blend-if move in order to get this to work.
I think you could get around that by creating a duplicate layer for blending, and then applying multiple luminosity blends/flattening to it instead of the background layer. When you're done, you can just do the blend-if moves on the duplicate layer once, then set the duplicate layer to luminosity mode. I just did a quick test and it seemed to work well enough.
I suppose the number of luminosity blends you can get away with depends on the noise in the channel in question.
For your final point, I usually do curves for additional contrast, too. I just think a luminosity blend usually gives a nice, quick improvement, resulting in a better starting point for the image.