A false profile (that is, a profile with a non-standard gamma value) can be expressed as pencil-drawn curves generated by Photoshop. While this has little practical relevance, it allows us to understand what assigning a false profile actually does to an image by simply inspecting the shape of a curve.
Some questions are dangerous, but some color profiles are curves.
A false profile (that is, a profile with a non-standard gamma value) can be expressed as pencil-drawn curves generated by Photoshop. While this has little practical relevance, it allows us to understand what assigning a false profile actually does to an image by simply inspecting the shape of a curve.
By Marco Olivotto from “Color Correction Cosmos series”
Dangerous Questions
The apparent simplicity of a question is often deceptive. If in doubt, try these:
Question #1:
Can you name the elderly lady who lives above the bakery located at number 52 in Wye Close, Porton, Wiltshire, United Kingdom?
Question #2:
What is the meaning of life?
An alternate way to state question #2 is in the caption of the following figure.
An original, two false profiles. Which is which? Which is true? How many can there be?
It takes 22 words to state the first question and just 6 for the second. The first sounds like an impossible challenge – and indeed it may take some time to work out an answer. But there’s a failsafe method to get there:
- Bring yourself to England.
- Drive to Wiltshire.
- Find Porton.
- Find the Close.
- Find the house.
- Ask the baker.
“Guinevere Catherine Rodgers.” Solved. Nevermind.
Let’s see where question #2 may lead:
- Silence.
- Silence.
- […]
- Silence.
“Dan Margulis, we have a problem.”
Dangerous Friends
Not quite your regular curve. Believe me.
Let me state once and for all that Davide Barranca, aka DB, isn’t a very nice guy and belongs to a very dangerous category of human beings. The main problem is that his questions sound like #1, yet they hide the complexity of #2. He recently asked one, in his usual gentle way: “do you have an idea why…?”
My first reaction was more or less along the way of “Piece of cake!”. The result: two weeks later I was staring at a crack in the wall in front of me and the cake was completely rotten. Yes, I did cry a bit, too – since I you ask. The actual question was: could you tell why a low-gamma false profile in Multiply mode usually works well, while a high-gamma false profile in Screen mode looks horrible? Frankly, he didn’t actually say “looks horrible” – he was a lot more explicit, but at least you got the idea.
DB, aka Davide Barranca, has a good habit, though: when you’re stuck he’ll often send you a link to some article on some obscure website. These e-mails seemingly come out of nowhere and while you won’t immediately understand what he has in mind, if you’re patient enough you will often discover that there is a tiny light hidden in them. Follow it patiently, and it may turn into something which remotely resembles an embrionic answer to the original question. Many mumblings and a few curses later, the answer finally rears its head – not necessarily an ugly one.
I like coincidences, and here’s one. The link, first: click here. This article by Martin Evening was published on September 5th, 2007. I didn’t notice until I started writing this – and today, as I write this, is September 5th, 2011. Maybe I’m on the right track? The article ranks as “difficult”, I’m afraid, but the basic idea is this: you’re certainly aware of Blend Modes and you’re certainly aware of Curves. Well, they are the same thing: you can mimic (almost) any Blend Mode with a curve. And if you’re concerned about how close a curve may bring you to a Blend Mode, the answer is easy and clear: the two results are identical by whatever sensible standard.
If you’re interested in how this can be done, go and read the article – the recipe is in there. Good luck, will I add – for it’s not, I repeat, not for the fraidy-cat. But after wondering why DB had sent me into the forest, I started seeing a thin line through the trees. And I thought – if it can be done with Blend Modes then it can be done with something else, possibly?
Enough of this. Let’s dive straight in, sharks and all. Take a look at these pictures.
A Profile is a Curve
Left: the original (sRGB) – Center: sRGB, gamma 1.5 – Right: a curve emulating the false profile.
The picture on the left is a crop from a photograph which I’ve chosen because it has both very bright highlights and very deep shadows. The original is tagged sRGB, hence it has a 2.2 gamma. At the center, the same picture with a false profile, sRGB, 1.5 gamma, and converted back to standard sRGB. On the right, the original treated with a curve which I built in an attempt to emulate the false profile.
The effect of the false profile is, of course, to lighten up the picture. In case this is not clear you should go and read the original article by DB.
If you’re curious which curve I’ve used, you’ve seen it in the second image of this post. Notice that the curve is not a regular Bezier curve, but something computer-generated which is loaded as a preset of a pencil-drawn curve (notice the small selected icon in the top left corner). I don’t think there’s a way on Earth to exactly match this curve with a Bezier shape especially because, if you look carefully, the very first segment of the curve in the deep shadows is linear. The curve you see is indeed made by two different curves aptly joined. The reasons for this are complex and we’ll discuss them later, but for now we can say this has a somewhat serious impact on the behaviour of false profiles in the shadows – and not just the shadows.
Can you see a difference between the central and rightmost picture? I can’t, but maybe the differences are too tiny to be visible. Photoshop can help us, in these cases: what we need to do is make a two-layer document, overlay the two pictures we want to check and put the topmost layer in Difference Blend Mode. The result – and probably the less interesting image ever to appear in the RBG Blog – is this:
The difference between the rightmost and central image.
This is, by my standards, a black image. Is it perfectly black – that is, is it 0R 0G 0B in each pixel? Not exactly: there is a deviation, because you can easily measure random shifts of up to 2 RGB points in each channel. Yet this difference can’t keep me from stating that the two original images are de facto identical: to give you an idea, if you create a completely black artificial layer and apply Filter -> Noise -> Add Noise… at 1% you’ll have a greater deviation from the original than this.
The numerical difference between the two files is not due to an error in the process, but to rounding errors. The file we’re working on is 8-bit: any operation, in principle, can cause a rounding error in the least significant bit, and we need two operations to obtain the central image: first, a false profile is assigned; second, the image is converted back to standard sRGB. One operation only is required to obtain the rightmost image: a Curve. Assigning and converting, as well as curving, can induce errors. If you don’t believe it, try this: start from an sRGB image; duplicate it; convert the duplicate to Adobe RGB; convert it back to sRGB; take the difference of the resulting image and the original one. The result is of course black, but you can already see a deviation of 1 RGB point in each channel, randomly. Calling that “a difference” implies that we’re relying purely on mathematics. From a visual point of view, though, it would be awkward to say that the doubly converted version shows a noticeable deviation from the original, simply because no human eye could ever see it.
Conclusion: by whatever sensible standard, the two images above (central and rightmost) are visually identical. You can try this on 10, 50, 2000 images: if the result is constant (and it is, believe me) then the only possible conclusion is that the curve used to produce the rightmost image is a perfect emulation of the assignment of a false profile.
But is it Useful?
The question is, of course, whether there is any advantage in making a complex curve and use it as a substitute of a false profile. The answer is: in practice, no. But, as we’ll see in the next parts of this article, there is a serious theoretical advantage: we can mimic very complex operations with curves and have the immediate visual feedback of what these operations do.
Consider also that in practice you would need a curve preset for:
- every profile you’re planning to use;
- every false gamma you’re planning to use with each profile.
In other words, if your scope is to convert sRGB and Adobe RGB files from their original gamma to 1.5 and 3.0 gamma, you would need four different presets: one for sRGB going from 2.2 gamma to 1.5, one for Adobe RGB going from 2.2 gamma to 1.5, and so on. A panel like the one written by Giuliana Abbiati contains no less than 54 presets. These would need, in turn, 54 curves presets to be matched.
So, what we need to do is simply learn to visualize what complex operations involving false profiles will do: and we can do this simply by looking at the shape of a curve.
As a final word, the most important message in this article is that the United Kingdom certainly exists, and so do Wiltshire and Porton. While Porton may not be the best place for an elderly lady to live, I admittedly have no idea whether Wye Close is a real street name in the village, nor I know if there’s a bakery located at number 52. But, were it so, I would be totally astonished to discover that GCR lives above it.
[To be continued.]
Marco Olivotto is a Master in Color Correction since when he decides to spread the word about color correction techniques in Photoshop, still relatively unknown in Italy. Starting from 2011 he takes the Color Correction Campus, full-immersion one or two-day courses were the students work on the images applying the techniques taught in the class. The CCC soon gained a big reputation and plenty of enthusiasts followers. Marco also teaches colour correction techniques in important workshops and conferences. He is a Teacher-in-a-Box instructor and writes for the magazine Fotografia Reflex which hosts a monthly space about color correction.
Last but not least. Marco “The Voice” Olivotto is the author of our best tutorials.
The great Dan Margulis, who invented color correction in Photoshop, and was the first mentor of Marco publicly called him “a renaissance man” because of his eclecticism. Visit he site here >
