White point and matrix device input profile?

[Stephen T]

Hello,

A matrix profile is simply a linear transform from device RGB to linear XYZ.

However, there also is a wtpt tag in the ICC profile from Argyll CMS. Where does that come from?

Is the profile wtpt tag used to convert from profile white point to the white point of the PCS (a second computation step after applying the matrix perhaps)?

I would really like some guidance on what to do about the white point for matrix profiles.


Cheers, Stephen.

[Graeme Gill]

Stephen T wrote:

Hi,

Post by Stephen T
A matrix profile is simply a linear transform from device RGB to linear XYZ.

Mathematically it's not linear, since the per channel (gamma) curves are typically
a power like function.

Post by Stephen T
However, there also is a wtpt tag in the ICC profile from Argyll CMS. Where does that
come from?

ICC profiles are stored as PCS white (D50) normalized. The white point tag
represents the native white point of the device. (Not true for V4 display
profiles though, which force the white point tag to be D50 as well).

Post by Stephen T
Is the profile wtpt tag used to convert from profile white point to the white point of
the PCS (a second computation step after applying the matrix perhaps)?

I'd say "device white point to PCS white point".

Graeme Gill.

[Stephen T]

OK folks,

Here are my findings for linear matrix camera input profiles:

1) For -am profiles and linear device RGB, the wtpt and bkpt tags are irrelevant in practise. However, we still might check the colour of the estimated wtpt (xyZ) to confirm it is near to D50, but don't worry if L* is not 100.

2) The rXYZ, gXYZ, bXYZ define the matrix transform from linear device RGB to D50 XYZ. The magnitudes of the rXYZ, gXYZ, bXYZ depend on any optional scaling that can be applied to prevent clipping.

3) Normally, we scale linear device RGB so that the white patch = 100 (maximum device response). This can be done in the raw photo software (linear scaling) or in Argyll CMS using -u (automatic scaling) or -U (manual scaling). Both raw photo and Argyll CMS scaling are equivalent. I suggest to use manual scaling in Argyll CMS and then check that no individual RGB exceeds 100 in an xicclu -g - fb plot.

Happy profiling.


On Wednesday, 30 July 2014 10:38 PM, Graeme Gill <graeme-***@public.gmane.org> wrote:



Stephen T wrote:

Hi,

Post by Stephen T
A matrix profile is simply a linear transform from device RGB to linear XYZ.

Mathematically it's not linear, since the per channel (gamma) curves are typically
a power like function.

Post by Stephen T
However, there also is a wtpt tag in the ICC profile from Argyll CMS. Where does that
come from?

ICC profiles are stored as PCS white (D50) normalized. The white point tag
represents the native white point of the device. (Not true for V4 display
profiles though, which force the white point tag to be D50 as well).

Post by Stephen T
Is the profile wtpt tag used to convert from profile white point to the white point of
the PCS (a second computation step after applying the matrix perhaps)?

I'd say "device white point to PCS white point".

Graeme Gill.