What tone curve does Photoshop use for sRGB?

[alexit3748321]

Hi,

Simple question really... what tone curve does PS apply when displaying sRGB images? Is it the piecewise sRGB encoding function or a pure 2.2 power curve?

[D Fosse]

First of all, the premise for the question is faulty. Any tone curve will be remapped through standard color management, so that the net visual result on screen is linear. It's a standard profile conversion from the document profile into the monitor profile. Whether it's sRGB, gamma 2.2 Adobe RGB, or gamma 1.8 ProPhoto, it will all be displayed correctly on screen. You don't "see" the document profile tone curve.

Without color management, however, the document's native encoded tone curve is sent directly to screen uncorrected.

The sRGB specification is a custom and irregular tone curve, not a regular gamma function. Among other things, it has a flat "toe" section near black. The reason for this is that sRGB was originally specified as a description of a CRT monitor.  It was, in a sense, the seed for modern color management.

The sRGB tone curve is also used in Apple's Display P3, and Photoshop's Image P3 (those two are identical).

Modern LCD panels don't behave quite like the CRT monitors did, and so it's a bit less useful. But it has been close enough to survive to this day.

[alexit3748321]

No, it's not a faulty premise. The question is whether Photoshop applies a piecewise inverse EOTF or a 2.2 power forward EOTF. Which transfer function does Photoshop use for display output?

SRGB uses an encoding function with an offset 2.4 power with a linear section, however the EOTF is a pure 2.2 power function. This, by the way, is exactly how Apple handles it in both sRGB and Display P3. You can find a good discussion here with references to the actual standard.

sRGB piece-wise EOTF vs pure gamma - Discussions - ACESNext / VWG – Output Transforms - Community - ...

[D Fosse]

Neither, you still misunderstand. Photoshop applies your monitor profile.

Standard color management remaps the tone curve from the document color space to the monitor color space.

Are you using a calibrator to make your monitor profile? A calibrator measures your display and builds an icc monitor profile based on that measurement. The transfer function in the profile is whatever your monitor does.

[alexit3748321]

Sorry, but that makes no sense. You can't just ignore the encoding and aåply the inverse of whatever your monitor transfer function is.

[D Fosse]

@alexit3748321 

Yes, that's just what you can. Of course it's not "ignored" - it's corrected for.

I see you're coming from aces/ocio, and as I already said, the big difference is that icc-based color management uses a monitor profile based on actual measurement of the display.  Aces/ocio doesn't do that, so the concept may be unfamiliar.

I suggest you read up on display color management. My co-poster here, Conrad Chavez, has written some excellent stuff, maybe he can point you to something.

[alexit3748321]

And the question is how the this is "corrected" for. Say you have a pure 2.2 power display displaying an sRGB encoded image. What, if any, output display transforms are applied? The right way to handle it is to apply no output display transform - you're already correctly set. ACES, for example, applies an inverse of the sRGB piecewise function, which is incorrect.

With ICC color management you have corrections for the display response curve but it's the same principle that applies. If my display is set to a pure 2.4 power curve the color management system should apply a 2.4->2.2 correction.

Your assumption would be correct if it's simply about converting the image to linear light. Say, for example, if the image is encoded in Prophoto RGB with a pure 1.8 power curve, then you simply linearize it in accorande with your display profile. This, however, is not how sRGB should be treated, which is also the point of my previous link. An sRGB image treated this way will have the toe lifted, which is incorrect.

[D Fosse]

BTW - since ACES is mentioned here, some clarification. ACES-based color management (OCIO) is used in the film/video industry. It operates on exactly the same principles as icc-based color management - but with one important distinction: there is no measurement of the display, no monitor profile. The screen output is just defined as a range of preset standards.

https://helpx.adobe.com/after-effects/using/opencolorio-aces-color-management.html 

In terms of accuracy, icc based color management takes it one step further by using a monitor profile. It makes sense to not do that in video, since a full monitor profile conversion for each frame may not be very practical.

ACES-based color management does not do a full remapping. So there you need to match file and display as closely as possible beforehand, and any variation from the preset tone curve will be visible. With icc-based color management that is not necessary, full remapping is done in real time.

[Conrad_C]

My level of expertise in video color management seems to be well below most of the participants in the thread you linked to. ACES is a video production workflow, and a lot of the people who post here are web- or print-oriented, including me, and aren’t familiar with ACES. But I’ll give it a whirl anyway. Naturally, I’m not 100% confident in the following answer.

It sounds like you are interested in the Photoshop angle to the first post in the thread: Which sRGB EOTF encoding is better. You want to know which of those two Photoshop is using, right?

First, I’m assuming EOTF (electro-optical transfer function) is the same as TRC (tonal response curve). As far as I understand all this, I am guessing a Photoshop sRGB image would use option #1: Piecewise.

How I got to that conclusion: Photoshop doesn’t really set the TRC on its own, it hasn’t typically done that since ICC-based color management became common a couple of decades ago. After that change, the TRC of a Photoshop document became set by the ICC color profile embedded in it. That’s why I think that the Photoshop color values you’re asking about are the ones Photoshop sends to the OS for display, which are determined by the document’s color profile. (If the document has an embedded color profile, it’s that; if the document doesn’t have an embedded profile then it’s the Photoshop working space profile for that color mode.) The OS interprets the received document color values using the document’s color profile, then the OS runs that through the display color profile, creating the tones and colors you see on the display.

For example, if a Photoshop document’s color profile is sRGB IEC 61966-2.1.icc, then I am guessing the colors are encoded using the TRC/EOTF in that profile, which is the piecewise method (with that toe in the shadows). If a document uses a different RGB profile, that’s probably a pure gamma curve, because of the next paragraph…

Regarding option #2 in the first post of the linked thread, an “image with a pure power sRGB Inverse EOTF.” I don’t think it’s common for Photoshop to produce that exact thing unless forced to, because of two things. The sRGB profiles I see provided in the macOS and Adobe default color profile folders use the piecewise EOTF, if I’m correct that is what the “parametric tone response curve” value is (figure A, notice how both say 1024 points). And, although you could also use Photoshop to apply a profile that does have a pure power EOTF, and there are a lot of those, none of those profiles are sRGB (figure B, notice how all state it as a gamma value). Because sRGB and pure power EOTFs tend to not exist in the same ICC profile, in typical Photoshop usage it might be rare or nonexistent for you to see images with both sRGB primaries and a pure power EOTF.

The following profile info screen shots are from Apple ColorSync Utility. There may be better tools, but this one comes free with my computer.

Figure A:

Figure B:

I think the first poster asked about option #2 (pure power) because there may be some video/3D applications that don’t use ICC profiles, and might apply that type of “sRGB primaries but not the weird sRGB TRC” encoding when the user targets certain displays used for video production. Or they do support ICC profiles but have the option not to. But again, I’m getting out of my depth on that one…

You said “display output” in the question, and I have to reiterate something D Fosse was getting at that is hopefully clear from what I just wrote: Photoshop is not directly involved in sending color values to the display. It sends them to ICC-based system color management, which then runs them through the ICC display profile selected for the OS. So if you are actually asking about the color values received by the graphics hardware, you want to look at the TRC/EOTF in the display profile selected in the OS, not at Photoshop.

The question is arcane enough that I’m open to learning from anyone who can point out where these conclusions aren’t correct. Even if this is not totally correct, I hope this at least helps figure out what happens with Photoshop.