Re: PQ HDR in PNG - draft review

Ah, my mistake.
What we have implemented is support for parametric curves, which I had
incorrectly assumed was ICC V2.
Anything that can't be described with parametric curves, we use a 3D-LUT
for, which is both less accurate and less efficient.


On Sat, May 20, 2017 at 7:26 PM, Lars Borg <> wrote:

> For chrome to do  color management efficiently, everything needs to be
> done on the GPU.
> Tables are generally much less efficient than doing using math on the GPU,
> and my impression
> is that ICC v4 relies much more heavily on tables than ICC v2.
> Yes, I see the point, but I don’t see there being more LUTs in V4 profiles
> than in V2, rather the opposite.
> With v4 we can use the parametric curves instead of the sampled 1D LUTs in
> V2.
> For example the standard sRGB profile  (colorimetric intent) can be
> created entirely without LUTs in V4, but requires LUTs in V2.
> Ditto display profiles.
> Lars

Received on Sunday, 21 May 2017 05:56:41 UTC