[argyllcms] Re: Camera calibration: LUT only as good as matrix?
- From: Torsten Bronger <bronger@xxxxxxxxxxxxxxxxxxxxx>
- To: argyllcms@xxxxxxxxxxxxx
- Date: Wed, 03 Jul 2013 21:48:46 +0200
Hallöchen!
Maciej Bryński writes:
>> 3. Create a profile:
>>
>> colprof -am -u DSC02310
>
> I have best results creating single shaper + matrix profiles.
> Try that.
To be honest, I don't think that anything beyond matrix-only makes
much sense. For one thing, sensors are pretty much linear devices.
And then, the measurement accuracy is not high enough in my opinion.
The camera is not a standard observer, so its colour "sensations"
are only close to what is described by XYZ, Lab, or whatever human
colour space. It is like two 2D planes in 3D space that have only
the origin point in common. Surely (hopefully), they are quite
close to each other, but it also means that a real match is
impossible, no matter how many patches you measure and how well your
fit goes.
I would be interested though in papers that have actually estimated
these error sources.
Tschö,
Torsten.
--
Torsten Bronger Jabber ID: torsten.bronger@xxxxxxxxxxxxxxxxxxxxx
or http://bronger-jmp.appspot.com
Other related posts:
