[argyllcms] Re: White Point

  • From: Elle Stone <ellestone@xxxxxxxxxxxxxxxxxxxx>
  • To: argyllcms@xxxxxxxxxxxxx
  • Date: Mon, 21 Sep 2015 06:31:06 -0400

On 09/20/2015 10:11 PM, Ben Goren wrote:

On Sep 20, 2015, at 5:15 PM, Hening Bettermann <hein@xxxxxxxxxxxxx> wrote:
I have earlier used -am, but -aX gives better Delta-E values.

But if you're making a profile from a commercially-available reflective chart,
-as -qm is your best bet.

Assuming the raw processor is outputting interpolated target chart data without applying its own TRC to the data, then "-as" assumes that the sensor (or sensor + internal camera "pre-processing" before writing the raw file) responds nonlinearly and also differently in each channel.

Is there a reason to think Hening's camera is saving significantly nonlinear results in the raw file?

Are there enough points on the target chart to accurately model any presumed non-linear response of the "camera plus internal preprocessing"?

Some raw processors (digiKam's raw processor) do apply a tone curve over which the user has no control, in which case "-aS -qm" or "-aG" would seem to be a better choice.

Low deltas look really nice as they are printed to the terminal. But I think it's pretty important to also plot the xicclu graph and ask if the resulting curves make sense.


Other related posts: