[argyllcms] Re: Modifying internal display LUTs
- From: Graeme Gill <graeme@xxxxxxxxxxxxx>
- To: argyllcms@xxxxxxxxxxxxx
- Date: Fri, 28 Oct 2011 13:47:20 +1100
Knut Inge wrote:
> This has recently been brought to my attention. Evidently, some monitors may
> be able to change their LUTs even though it is not documented. Does anyone
> know links to such discussions anywhere?
There are no technical discussions about this as far as I am aware, as
such things are all proprietary.
> So the advantage of doing anything in the display would be that it is
> probably 10, 12 or 14 bits with direct access to the panel with a minimum of
> quantization in-between. Compared to the (usually) 8-bit quantization that
> happens when crossing the DVI link.
There is no guarantee as to what internal depth a display supports, or
how it is implemented (the native LCD often has limited depth, and extra
depth is faked using spatial dithering). Most displays probably don't
have accessible internal LUTs - they have some sort of firmware/hardware input
to
sRGB like conversion. Only high end graphics/color displays tend to have
accessible internal LUTs, and there are no standards and none of it
is public. Note also that analog VGA and Display Port allows more than 8 bit
precision, and even DVI allows it in theory, using dual links (in practice
I suspect this has never actually been implemented by any display or graphics
card.)
> But any LUT (especially 3D LUT) can only decrease the gamut, right?
Yes.
> The reason would be to calibrate the behaviour to some standard response. If
> you
> can characterize the GPU buffer -> visible light, it seems that in most
> cases one would prefer to give the software color management component full
> freedom and full access to the capabilities of your display to do the
> optimal mapping from image file to visible light?
That's my general opinion, although there are circumstances when
another arrangement may be preferable, such as feeding video into
the display, since few video systems support color profiles.
Please note the distinction between calibration and profiling though.
Given limited precision of control, and the way ICC profiles deal
with the white point and brightness (both are generally normalized
out of ICC profiles), then calibration has it's place.
> -For day-to-day use it would make a lot of sense to upload a "perfect" sRGB
> correction curve into the display, meaning that non-color-managed
> applications would look as good as possible. Ideally, when loading e.g.
> photoshop you would want it to upload a new/swap display LUT into something
> more "native"
If you have a wide gamut display though, you are then unable to make use of it.
Doing the color management in the computer system has the advantage that
different color spaces can be mixed on screen (although they may visually
interfere with each other). One problem is that until recently, windowing system
visual elements have not been color managed. I understand this has been fixed
in the latest versions of OS X and MSWin, but not in Linux (although there are
some workarounds for Linux available).
Graeme Gill.
Other related posts:
