Question about convert to profile

[J Isner]

I have a Photoshop document with embedded profile sRGB. The document is "grayscale."  That is, R=G=B for all pixels. When I convert the document to AdobeRGB, I was surprised to see the RGB numbers change. For example, (23,23,23) becomes (29,29,29). The L of Lab remains the same, which indicates the tone itself has not changed. I thought grayscale RGB numbers are the same in sRGB and AdobeRGB, and that only non-grayscale numbers are different. Was I wrong to think this?

Here are my Color settings

[Stephen_A_Marsh · Accepted solution]

Adobe RGB has a 2.2 gamma tone response curve.

sRGB has a custom TRC, close to 2.2, but not the same near the deep shadows.

Testing two profiles with exactly the same TRC would help to isolate if this is the issue.

[J Isner]

That's interesting, and it answers my question. Thanks!

[D Fosse]

Yes, numbers will change when you convert to a different profile. That's the remapping into the target color space in order to maintain visual appearance.

In other words, this is normal and expected.

Profile conversions go through a reference color space (profile connection space), either Lab or XYZ. So Lab numbers will stay the same throughout (unless there's gamut clipping).

[J Isner]

My question only concerned grayscale tones, and I still believe they should not change when converting between color spaces with the same gamma  I think Stephen identified the problem: it's the nonlinear gamma of sRGB.

[D Fosse]

But few color spaces actually have the same tone response curve (gamma), so in practice you will always get this shift in numbers.

The sRGB curve is highly idiosyncratic and, especially in the shadows, very different from the regular gamma 2.2 curve in Adobe RGB. The sRGB curve is very steep in the shadows and with a flat "toe" near black.

In fact none of the standard working RGB spaces have the same tone response curve. ProPhoto is gamma 1.8. And then of course you have all the grayscale profiles, all with very different tone curves and shifting numbers.