[argyllcms] Re: help with camera profile
- From: Stephen T <stwebvanuatu@xxxxxxxxxxxx>
- To: "argyllcms@xxxxxxxxxxxxx" <argyllcms@xxxxxxxxxxxxx>
- Date: Mon, 30 Jul 2012 16:52:09 -0700 (PDT)
Alberto, you need an independent, objective test.
Stop worrying about profiling errors and wasting time evaluating results on
screen.
Get a ColorChecker 24, photograph that, process as linear 16-bit TIFF (exactly
as you did when profiling), evaluate delta-E 2000 for the CC24 with profcheck
and your IT8.7 profile.
This test is independent of your IT8.7/2 target, workflow, RGB colour spaces,
display, printing, personal biases, etc. Delta-E 2000 is an objective,
quantitative measure of colour difference and reasonably consistent with human
perception.
Beware that you will have to match exposures, i.e. match the linear RGB (.ti3
output from scanin) for a neutral middle-grey patch (L* =
50) in both the profiling and test photos. Otherwise, you could have large
delta Es simply because of L* differences.
The illuminant might be different between the two shots, but we usually expect
(assume) camera profiles can be used with similar light sources (e.g. a D50
profile should provide "satisfactory" results for D65).
ST.
________________________________
From: "alberto.ferrante@xxxxxxxxxx" <alberto.ferrante@xxxxxxxxxx>
To: "argyllcms@xxxxxxxxxxxxx" <argyllcms@xxxxxxxxxxxxx>
Sent: Monday, 30 July 2012 7:21 PM
Subject: [argyllcms] Re: help with camera profile
As usual, I forgot the subject in my previous post... I apologize for
double posting.
I checked for clipping and, no, those patches are not
clipping while
generating the profile (i.e., during the conversion
from raw of the
target for processing): the L15 patch is at about 210,
210, and 15 for
R, G, and B channels, respectively. They may clip,
along with other
patches, once the final profile and curves are
applied, but this is a
different story.
I re-shoot the target with the
7D in the same
conditions of the 5DIII shoot (flash light) and the
situation improved.
I now have only one patch (the usual L15) with an
error of 22. If I
remove it, I get the peak error below 15 (14.
something). Regardless the
lightning conditions, though, the L15 (and
K15) patches are the ones
exhibiting with higher errors. Though, when
generating LUT-based
profiles, errors are much lower (around 2-3 as a
peak).
I tried
changing the output colorspace to Prophoto, but this
only slightly
improves the situation (the peak error goes to 20
instead of 22).
Regards,
Alberto
Iliah Borg wrote:
>> If I try
to do the same
with the 7D profile, the error gets higher (27
>> as
peak) and even
after removing 3 patches it stays to 17. The "bad"
>>
patches are
always orange/yellow.
>>
>it might depend on the output
colour space
and how they get their. I mean, have
>you checked for
clipping?
Invita i tuoi amici e Tiscali ti premia! Il consiglio di un amico vale più di
uno spot in TV. Per ogni nuovo abbonato 30 € di premio per te e per lui! Un
amico al mese e parli e navighi sempre gratis: http://freelosophy.tiscali.it/
Other related posts:
