does the Adobe RGB specification intentionally specify 307/256 gamma?

[Jaldebehot]

This is a very minor thing. But I was reading the Adobe RGB document here, and I was wondering if it's intentionally supposed to have a specific quantized hex value 0233 for the gamma rather than an ideal 2.2 - similar to how the sRGB specification also asks for a specific rounding.

I also notice that the Adobe RGB spec has specific instructions for creating a v4 profile compared to a v2 profile (different black point for example). I was curious what your thoughts are on this, since I notice you mention the v4 a98c profile has a more precise gamma value closer to decimal 2.2.

[saucecontrol]

That's a good question...

The spec is very clear that the value is to be precisely 0x0233. However, it's defining the color space in terms of the ICC v2 profile values with all the rounding included, while still hinting at what the intended values are.

Another hint to the intent is the section on tolerances for calibrated systems, which says

The gamma value should be within the range 2.190 to 2.210 for the individual color components and for the neutral axis.

So they've defined a range that's to be considered 'good enough', which happens to be centered at exactly 2.2.

In practical terms, the difference doesn't matter when creating a V2 profile, because 0x0233 is the closest value to 2.2 that can be encoded. But the profile does nothing on its own -- it's merely an input to your color management software. It turns out that many of the software implementations understand that 0x0233 is just an approximation for 2.2 and will 'upgrade' the value on reading the profile. The same actually happens with sRGB, with most higher-quality software recognizing anything within a certain tolerance of sRGB and upgrading it to the actual sRGB definition internally.

So what about V4? Well, given that a lot software will recognize Adobe RGB from its profile and upgrade it to a more precise definition, the only value that a V4 profile can add is for software that takes the profile values as they are. In that case, something that gets you closer to the ideal is adding value. In other words, a profile that keeps the V2 value may as well be a V2 profile.

Similarly, the slope limiting function mentioned in Annex C of the spec is implemented in most higher-end software, but I've included it explicitly in my V4 profile so that if taken literally, the profile will give a better result.

[Jaldebehot]

That makes sense. Thanks for that explanation. :)