[argyllcms] Re: Brainstorming about profiling Android devices

  • From: Graeme Gill <graeme@xxxxxxxxxxxxx>
  • To: argyllcms@xxxxxxxxxxxxx
  • Date: Thu, 15 Mar 2012 08:54:09 +1100

János, Tóth F. wrote:
> I tried to create a curves+matrix profile for a Nokia 5800XM display
> from 12 patches (100% RGB + 9 gray steps between 100% and 0%) and
> convert my main screen background image with Photoshop CS5.

You would need to make sure that it is either -aG or -aS (single
curve/gamma), and gamma may be more reliable than shaper curves.
If you added 50% color steps then you may get away with
-ag. It would be interesting to know the fitting error.

> Both were useless on the phone display. The skin-tone got better but
> there was an awful amount of posterization/banding on both images (the
> absolute intent was obviously useless but the other was also too
> damaged, it didn't worth it).
> (I guess that was a 6-bit display and the images were dithered and
> saved to 8 bit, so even the dithering noise was distorted by the
> internal quantization...)

It's hard to equate 6 bit with the level of dithering you describe.
Could it have been posterization due to a very inaccurate profile ?
(Typically 6 bit dithered is hard to visually distinguish from 8 bit.
 Many an Apple user is unaware that they are using 6 bit displays that
 purport to be 8 bit - I certainly don't notice it on my MacBook).

Graeme Gill.

Other related posts: