Re: Sum of three primaries < Monitor white?

Thomas,

First of all, I never wrote that the digital counts of three channels
are linear. And my additive assumption, well not mine, of those who
are working in color device characterization ( a ton of papers using
this assumption you can find easily in the Internet), is in fact not
that three digital counts are linear.


Then the sum of the individual outputs is not identical to the output of
the sums, case closed. [255,255,255] is truely the sum of [255,0,0],
[0,255,0] + [0,0,255], but that doesn't hold for intensities, as you
already found, because the mapping between the input (pixel values) and
the output (intensities) is nonlinear. Where's the problem understanding
that?

Second, Nop, case has not closed yet. Lets see. I am talking about the
highest digital counts for three channels. Do not get me wrong. This
means if I have the XYZ values of the lower digital levels such as
[200,0,0], [0, 30, 0], and [0, 0, 100], the sum of their XYZ is NOT
expected to be close to the measured XYZ of [200, 30, 100]. But... for
[250,0,0], [0 255,0], and [0,0,255] it is expected to be CLOSE (not
exactly equal) to the native monitor white.

Can you explain for me why the experiment of mouro in the same thread
shows that what I wrote is true?

OK, please do what I suggest you. Read some papers about monitor
characterization. Or you might want to read the section "Additivity"
in page 26 of this paper "Colorimetric characterization of three
computer displays (LCD and CRT ) by Jason E. Gibson, Mark D.
Fairchild .
http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.32.3742

This paper will show you what I wrote is true.


Thanks,

Regards,


On Oct 24, 2:04 am, Thomas Rrichter <t...@xxxxxxxxxxxxxxxxx> wrote:
colorresearc...@xxxxxxxxx schrieb:

Hi Thomas,
No one says that three channels are linear.



And the additive
assumption in model based method for monitor characterization is not
the assumption of linearity of three channels.

How do you feed the monitor then? Why do you expect that the red channel
gets the same intensity no matter whether you feed the monitor input
with red or with white? There is no reason why this should be the case.

That is why for model
based method we need to linearize three channels first and then
multiply the linear RGB with a transformation matrix to obtain the
predicted XYZ values. A

Monitors aren't controlled in RGB space, they are feed with signals in
the gamma-corrected R'G'B' space.

gain, I am pretty sure that for conventional
additive monitors,

Conventional monitors aren't additive - their intensities aren't linear
functions of their input signals.

the sum of the XYZ values of the three monitors
should close to the measured XYZ of native monitor white.

Why? This would be only the case if the intensity of a single channel,
feed with full intensity, is identical to the intensity of that channel
when the input lines receive a signal equivalent to full white. Neither
of that need to be the case. For example, for some monitors, you can
"adjust" the white-point. This wouldn't be possible if full intensity
input would be mapped to full-intensity output.

And the
experiment of Boscarol posted in the same thread shows you that this
assumption is true.

Perhap, you might want to read the presentation of Wyble at
http://www.cis.rit.edu/info/IA_S2003_talks/Wyble_talk.pdf
It might explain what I mean better, I think.

I think you still don't understand what you're measuring - your
underlying assumption is wrong, which is why you don't get the result
you expect. A monitor is by no means a linear device - you cannot assume
that the pixel values are in linear relation to the intensities.

So long,
        Thomas

.

Relevant Pages
Loading