[Openicc] Linux CM ideology, was: meta data in test chart

[Chris Murphy]

On Jan 28, 2011, at 3:06 AM, Richard Hughes wrote:

> 
> As large gamut display devices become more and more "mainstream" I
> think this kind of thing is going to be more and more important.

And another issue is that the vast majority of LCDs that aren't even attempting to be wide gamut, regularly depart from the sRGB blue primary. By a lot. So this causes all kinds of discrepancies between displays when it comes to blue hue. Some images with blue will appear more cyan blue and others appear very purple blue.

One thing that started to happen when Firefox turned on full display compensation (in betas) for untagged images, which were assumed to be sRGB, is that images with blue were rendered a LOT more purple because the sRGB blue is rather close to indigo rather than blue. It's based on CRT blue primaries, and the dyes to make the blue filters for LCDs typically just aren't that hue or that chromatic.

So there's a whole bunch of content on the internet that really isn't even sRGB anymore, we don't have a definable space for it. An argument could be made that we need an sRGB' that has a shifted blue primary that correlates better with LCD blue primaries, rather than non-existent CRT blue primaries. We aren't going to get LCDs to reproduce that blue anyway...

What ends up happening with simple matrix transformations without good perceptual rendering intents in display profiles (which effectively none of them have in terms of volume), is that sRGB blue has to be desaturated to get it to display on an LCD, and in the course of that simple desaturation done in the CMM, since the profile does not define it, we end up with purple blues. And people don't like it.



Chris Murphy