[Openicc] Manual profiling

[Chris Murphy]

a.) It can be useful or it can be useless depends on implementation.

b.) Helping the user set brightness and contrast is very useful. This isn't exactly trivial on LCDs, because first you need to determine which control to inform the user to manipulate: brightness, contrast, or backlight. Most displays don't have a backlight control, but if they do that's probably the one they should use. Most displays attached the brightness control to the backlight, although some attach that control an internal LUT. And most displays attached the contrast control to a LUT. First you need to figure this out for that particular display. Then have the user reset the display to defaults. Then proceed with manipulation of the single control that affects just the backlight.

c.) The primaries themselves probably just have to be accepted at face value from EDID, unless they are unrealistic values. There might be some value in a metric that validates the primaries anyway, and use some substitution if they're wildly out of line.

d.) For white point adjustments,  you're best off having the user look at features in their OSD (on-screen display) to put the display white point "in the ball park" rather than starting out at 6500K and knocking it down to 5000K with just video LUTs. Many, but maybe not most (I'm not sure) displays have internal LUTs to set the white point. Usually these are presets like 5000K, 6500K, 9300K, with them defaulting to 6500K. So if you've had the user to a display reset already, good bet it's already at 6500K. And if they want to change it substantially (close to 5000K) you may be better off having them find a way to change it in the OSD first, then use the manual calibration tool to fine tune that result. Otherwise they will get inferior results.

e.) It might make some sense to start a database of display devices for which manual calibration should gray out. These include displays that have high bit LUTs for fine calibration internally, rather than any dependency on the video card LUT.

I wish that VESA would hammer on vendors to support MCCS so that all of this could be made much more straightforward and standard without having to depend on the user to do certain things like access the OSD or even brightness controls. But presently, the manufacturers are all over the map on how their displays behave and are controlled. So it's a bit of a rabbit hole to get into.


Chris

On Apr 2, 2011, at 9:33 AM, Richard Hughes wrote:

> In gnome-color-manager I've got quite a bit of semi-working code to
> performing manual display profiling, i.e. asking the user to stare at
> the mid-gray box and get them to adjust sliders until the gamma looks
> correct. We ask them to do this 10 times, and interpolate the rest
> from that. This has never worked very well for me at all, and I've
> never managed to get a successful profile out of it. Even asking the
> user for the desired whitepoint and using the factory-present edid
> primaries seems to work better.
> 
> I'm now at the point of planning for GNOME 3.2, and I've not decided
> if I should "rip it out" or "spend lots of time on it to make it
> work".
> 
> Your advice please: Is manual profiling useful for people too poor for
> a hardware calibration device or is worse than useless?
> 
> Thanks,
> 
> Richard.
> _______________________________________________
> openicc mailing list
> openicc at lists.freedesktop.org
> http://lists.freedesktop.org/mailman/listinfo/openicc