The “False Profile / Multiply” technique avatar
The “False Profile / Multiply” technique

by Davide on 10 June 2011

When my dear friend and co-blogger Giuliana Abbiati coded a free Photoshop CS5 extension called “False Profiles” I suggested her to add some quick and dirt illustrated examples; then we started wondering that false profiles are a simple yet powerful tool, which may require more than a couple of pictures.

© Roberto Bigano - Plastic Girls (Heilderberg, Germany. Hauptstrasse)

Actually, the toolset is easy to use but the theory behind it is quite deep: Dan Margulis originally invented this technique some 10 years ago, and in his book “Professional Photoshop” 5th edition he gives the background to understand all the subtleties that may follow, plus some extra +450 pages worth of reading. There’s also an article (from Electronic Publishing magazine) he wrote on the subject that you may read online. I leave to Marco Olivotto the honour to divulgate/elaborate the theory better than I can, for he’s the actual teacher of the group.

I’ll simply show here some application of False Profiles for tonal corrections, combined with the recent add of the masked Multiply step (sounds like a super-hero): it’s used to equalize the luminosity of pictures that contain at the same time very bright and dark areas, and/or to recover detail in there.
I’ll be using pictures from a soon to be released Roberto Bigano’s photographic book called Plastic Girls; all the files are tagged sRGB so this will be the reference ICC throughout the post.

You can download the free Photoshop CS5 extension here, where Giuliana approaches the theory as well: I suggest you to read it.

1. False Profile in a nutshell

False Profiles (FP) are – not only, but also – a way to lighten or darken a picture, without actually changing pixel values – only their interpretation. Let me explain briefly.

If I show you a middle gray board, it’s… middle gray.
If I add a “White board” writing on it you may think “Uh, quite dark to be a white board”.
On the contrary, if the writing says “Black board” you will think “Uhm, it’s too light to be a black board”.
You see, feelings depend on your expectations.

Back to pictures: if an image is tagged with an sRGB profile (which has a gamma approx of 2.2), it’s… that image.
If I tag that image (Assign Profile) with a custom made sRGB profile with a modified gamma of 3.0 you will perceive it as quite dark.
If I tag (Assign Profile) with a custom made sRGB profile with a modified gamma of 1.4 you will perceive it as quite bright.
It’s the very same original image, pixels didn’t change, but the way Photoshop displays them changes.

(Fig. 1) Different gamma compared, original in the center

The whole story is a bit more elaborate than this and it involves also color primaries, but this will suffice in order for you to follow the examples. Download the free extension, which automatically installs all the needed false ICC profiles, and get ready.
(Doubts? Read Giuliana’s post.  Read Dan Margulis original article about False Profiles. Read Charles Poynton famous Gamma FAQ. Read Dan Margulis’ Professional Photoshop. Subscribe to ColorTheory group and ask the Maestro).

2. The False Profiles tecnique

A straight application of the False Profile (FP) idea: use a FP to brighten a dark picture (Fig. 2).

(Fig. 2) Original picture (left, too dark) and the application of a false profile (right, gamma 1.5)

The steps to follow:

  1. Assign a gamma 1.5 profile (pixels don’t change, only the way they’re displayed) clicking on Giuliana’s extension. Try different gamma, and pick up the one which gives you the more visually correct look.
  2. Convert to Profile (in Photoshop from the menu Edit – Convert to Profile…) back to sRGB. Remember to add this convertion step! This way the pixels are finally changed to their final value. You don’t want to give someone else (like a photolab, a printer) pictures tagged with a weird gamma profile, believe me.

So, Assign makes the picture look different just because it differently interprets the data; it’s up to Convert writes down the correct numbers on a “standard” profile.

3. The Multiply technique

Now let’s see the Multiplication (or Multiply) only in action. We need a picture with some critical highlight detail that we’d like to enhance (Fig. 3) You see in the first line the original and the final result, plus all the details to get there.

(Fig. 3) Standard multiplication steps explained in detail (no use of False Profiles)

Let’s review this time only all the necessary steps of the Multiplication.

  1. Duplicate the background layer and set its blending mode to Multiply.
  2. Add a layer mask, clicking the appropriate icon in the layer palette.
  3. Apply the RGB composite to it, with the Image – Apply Image command (we’ll use different channels later on).
  4. Run a Gaussian Blur filter to the layer mask (here 20px, it may be larger if you work with high res pictures).
  5. Optional: flatten and save.

If you notice, the highlight detail is enhanced. Setting a layer to the Multiply makes the picture darker (Multiply blending mode explained here), while adding a so-called luminosity mask as the layer mask means: please Photoshop, let the darkening to affect mostly the light areas of the picture, and quite less all the rest, almost nothing in the shadows. Mind you, any mask applied to a layer that changes dramatically the luminosity of what lies below it, must be blurred in order to increase the contrast (funny, try it), so here’s the why of step 4.

Variations on the Multiply theme

As a suggestion, instead of duplicating the background layer you can use a straight, unmodified Curves or Levels Adjustment Layer set to Multiply: it’s exactly the same.

And if you feel the effect may be stronger (i.e. more darkening needed) duplicate the multiply layer one more time, or as many time as you need. You can keep the layer mask, or as a slight variation clip the extra multiply layers on the first one (ALT+click in between the layers); or just keep duplicating Multiply layers, merge on the top the result (Command + Option + Shift + E on a Mac, CTRL + ALT + Shift + E on a PC) and apply to this single layer the mask that you like, trashing all the Multiply layers below. It doesn’t make a dramatic difference – except for the use of Adjustment Layers to keep the filesize small. It’s quite easy to give each option a try and use the one that we like the most.

4. Multiply + False Profile combined

So far we’ve seen that a False Profile can lighten a picture in a peculiar way, while the Multiplication step will make it selectively darker. What if we mix the two?

(Fig. 4) False Profile G1.8 plus three Multiply steps luminosity masked

What happens is that the picture’s tonal values get “equalized”: the three Multiplication layers alone make the picture too dark, even in the highlight that we’d like to recover. But highlight and shadow’s values get closer, so that a False Profile G1.8 can resuscitate the whole picture and the net result is more detail in both of the face’s sides. The exact order doesn’t matter too much, you can either:

  1. Duplicate the background layer, set it to Multiply and apply the RGB composite (aka Luminosity mask) as a layer mask, then duplicate this layer two more times
  2. Assign to the document a G1.8 profile

Or the reverse (Assign, Multiply). Merge on a separate layer with the usual shortcut (Command + Option + Shift + E on a Mac, CTRL + ALT + Shift + E on a PC) trashing the Multiply layers or simply flatten the result. Then Convert to Profile back to sRGB or the one of your choice.

We could be tempted to wonder that if a “Low Gamma profile + Multiply layer” works, also a “High Gamma profile + Screen layer” will.
No it doesn’t. If you don’t believe me try it yourself.
The why isn’t obvious, and lies in the different kind of darkening/lightening that a gamma curve and  Multiply/Screen blending modes apply. If you’re interested in the details, stay tuned for a RBGblog post will appear in the near future.

5. Mask options

In the last example as a layer mask I’ve used a luminosity mask, a blend of all three channels (see Fig. 3, last row). Actually, it’s really important to pick up the more appropriate channel, and from the more appropriate layer in order to tailor the effect to our needs.

(Fig. 5) First row: Original and False Profile G1.5 with two Multiply layers, masked with R channel or G channel. Second row, the R, G, B channels of the Original

As you see, the one masked with the R channel is lighter in the cyan jacket, for there the R channel is pitch black, while the face gets darker. The picture is more balanced using the G channel as a mask.

Not only the used channel is important, but also the source of the channel. In the box at left you see that it’s possible to choose between “Merged” and “Background” (plus all the existing layers, to which we don’t care). Being the Multiply layer darkening the picture, a channel of a merged version will be darker, and the Multiply effect weaker. On the contrary, a channel from the Background will be lighter, and the Multiply effect stronger. So, choosing from the Merged version leads to a lighter result (dark channel = less Multiply), while from the Background to a darker result (light channel = more Multiply). It may be a subtle effect, nonetheless if you start piling Multiply layers it becomes quite evident.

(Fig 5.1) Original in the center; left, lighter version (G 1.5, double Multiply layer masked with the B channel from the Merged version). Right, the same but with the mask coming from the Background.

6. More examples

Now that you master the principles, let’s play with Roberto’s images and see if we can apply the False Profile / Multiply technique proficiently.

(Fig 6.1) False Profile G1.8, triple Multiply (L-masked)

You see in Fig 6.1 that the multiply, in RGB, leads to moderate color saturation (here in the face), plus the usual boosted detail especially in the hat and the fabric.

(Fig 6.2) False Profile G1.8, Lab conversion and single Multiply layer L-masked while still in Lab mode

In Fig 6.2 I assigned a false profile G1.8, then I converted to Lab (Image – Mode – Lab). In Lab the Multiplication leads not only to an overall darkening, it also oversaturates the colors (dont’ ask me why, but a and b channels of Lab gets overlaid when the layer is multiplied). So I did the usual Multiply layer, masked with L from the background, in order to get a darker version. Blur the mask, convert to sRGB.

(Fig. 6.3) False Profile G1.8 and a double Multiply layer masked with R

Fig. 6.3 shows why choosing the right channel for the mask is important – the R, being light in the dress, makes the multiplication more effective just right there.

Well, being Roberto’s Plastic Girls pictures so refined, I guess the correction I’ve shown here are quite subtle: nevertheless the technique is powerful and this is just a showcase of some of the possibilities.
Happy photoshopping, and download Giuly’s False Profiles free Photoshop CS5 panel!

DB

Plastic Girls 1978-2010 is a photographic project by Roberto Bigano.

I have been shooting “Plastic Girls” for over thirty years. They fascinate me and I find they tell a lot about the culture of a country and the evolution of fashion and style.
According to a friend of mine, Beppe Maghenzani, I am an hermeneutic, Greek word difficult to translate but, in my case, means that every photo I take is complete in itself and holds myself in its entirety as well as the fact that I continue to work on this subject, forever.
It is very important to clarify that all mannequins, without exception, were photographed in the stores windows always through the glass and never with special permits. All these wonders are right there, available to anyone who has ability to see them.
Another thing that has always amazed me is the ability of window display stylists, not only to set up the store windows, but also to light, often in extraordinary ways, the objects.
I think every photographer has a lot to learn from them.

 

{ 4 comments }

JRP June 22, 2011 16:36

What is the advantage of this approach, compared to using adjusting the gamma in a masked, multiply blend mode Exposure adjustment layer?

Davide Davide June 23, 2011 09:23

Hello JRP,
you can fake False Profiles (pun intended :-) ) via Gamma correction in an Exposure adj. layer, it’s totally viable – I’m afraid I can’t see how it would be possible to merge the two step. Say that Exposure is lightening the image, if you set it to Multiply you’re reverting the effect.

As a side note, if you set to Multiply a 9,99 Gamma corrected Exposure layer, it’s close to multiplying the K channel (something that Dan Margulis often suggests).

Anyway, Marco Olivotto is currently writing a quite technical post about Gamma adjustments (explaining why low gamma + Multiply works, and high gamma + Screen doesn’t), it’ll be published soon, so come back!
Thanks

Davide

JRP September 3, 2011 11:09

“I’m afraid I can’t see how it would be possible to merge the two step. Say that Exposure is lightening the image, if you set it to Multiply you’re reverting the effect.”

Perhaps we are talking at cross purposes. The Exposure Adjustement provides for gamma correction too, so you can do the two in one.

marco marco September 6, 2011 16:35

There is a subtle but substantial difference between the gamma correction available in the Exposure Adjustment and the assignment of a false profile. One main point first: we should always think, when we make comparisons, to two steps: “assign false profile AND convert to original profile”, because assigning the false profile changes the appearance of the image without changing the numbers, whereas a gamma correction made through Exposure does change the numbers. In the first case, the numbers change only when the image is converted to some other profile, and in this case, since we’re bound to compare the results, the profile should be the original one.
Having said this, I need to explain how the gamma slider in the Exposure Adjustment works. It doesn’t really show gamma but a ratio of gammas: when set to 1.00 (default) it means “keep gamma as the original, whatever it is”. In practice: if you have an image tagged with a sRGB profile (whose gamma is 2.2), gamma = 1.00 in the slider means no change – hence, gamma 2.2. If the same image is converted to Apple RGB instead (remember that the gamma of this profile is 1.8) its apperance won’t change but gamma = 1.00 in the slider will mean gamma 1.8 – again, no change.
The interesting thing happens when you move the slider. If you set it to, say, 2.00, that means that you are miming the assignment of a profile whose gamma is (original_gamma / value_of_slider). In this case, you’re taking the original gamma and dividing it by a factor of 2. So, if you’re in sRGB or Adobe RGB, gamma = 2.00 actually means that you’re miming the assignment of a profile whose gamma is 2.2 / 2 = 1.1. If you’re in Apple RGB or ProPhoto RGB, the emulation is that of a profile whose gamma is 1.8 / 2 = 0.9.
If you find this confusing, it is. But it takes little time to get used to the idea.
Yet this is a simulation, and it’s not 100% correct. The difference lies and can be seen in the darkest shadows. Anything else is identical, and you can easily prove that by producing two images with the two techniques and overlaying them in Difference mode. Yet the shadows fail. This is because an RGB profile is linear at very small values. This is needed because otherwise the curve which corresponds to the false profile (a false profile IS a curve, indeed) would be too steep in the shadows, and you would experience a substantial amount of noise in those areas. The gamma slider doesn’t do that: it applies a curve, but the curve is NOT linear in the dark shadows. And in every case I’ve tried, the gamma slider is noisier in areas where there is a low value in any channel.
This may sound like a very innocent fact, but it is not. The keyword in the last sentence is “any”. If you have an area whose average color is, say, 230R 240G 5B, you’re dealing with an almost pure yellow: very bright, very saturated. But there is one channel which is very low, and it is of course the B channel. When you run the Exposure gamma on the image, the value in that channel may change very significantly, and you may experience a color shift in any pure or near-pure color you have. Invariably, a desaturation. If you’re playing with extreme gammas like 0.75, this may become a true problem: not only the shadows change, but a lot of other areas may.
You decide to use either method depending on the image. If all you need is to extract as much detail as possible from the shadows, go Exposure – but be warned: you’ll have to lick noise off your screen in the deep shadows :-) .
As DB suggested, I am working on an article. This is one of the concepts which I will discuss, so stay tuned if you’re interested.
Best regards!
MO

{ 4 trackbacks }

Previous post:

Next post: