Re: Colorimetric Accuracy in the Field
Re: Colorimetric Accuracy in the Field
- Subject: Re: Colorimetric Accuracy in the Field
- From: email@hidden
- Date: Tue, 04 Jun 2013 22:45:12 +0100
HelloI hope this gets through. Im not sure how to reply to the list.
Interesting thread.
Define colormetric accuracy – Colour in equals colour out?
Colour in depends on illumination, spectral response of surface (we
will take surfaces as example) and an observer, in this case 2˙
standard.(Not forgetting this also depends on instrumentation accuracy
– traceability to national standards – and repeatability).And
as Jeffrey suggests are we all standard 2˙? but thats another
debate. I think there is a new research on redefing the standard
observer for different age groups and
using sharpened cone responses to produce a new set of CMFs.
We can now define our subject in terms of colormetric input. ie we
have measured our Colorchecker SG ten times and went outside taken a
pic noting the SPD of the daylight using your Minolta
CS1000 exactly at the time of exposure.
And this is the bit I think is in question. Is the scene captured
by the camera an accurate colormetric representation? This depends on
how the device values are converted to CIEXYZ. This is the bit that
make everything colormetric accurate or not.
There are two main methods for calculating the transforms (camera
matrix) to produce scene colorimetry estimates from sensor data. The
first is a target based method (using Macbeth Color Checker for
instance) and the second is to use spectral colours produced by a
monochromator (the latter my preferred method).
Currently the approach used by both methods for error minimization is
to calculate the RMS error in CIEXYZ, CIE L*a*b*, CIE L* u* v* or in a
linear or non-linear based RGB space (ITU-RBT.709, RIMM RGB).
It is this matrix and how it is made and its performance which will
define wether you have colormetric accuracy. And as the camera CMFs
are not exactly the same as standard 2˙ CMFs there will be errors.
It is not easy to test wether the matrix you have used in your raw
converter is accuarte or not with out a lot of meassuring, shooting
and comparing values. But you have the rendered image of the
colorchecker SG and you can open an save in prophoto and convert back
to LAB and compare measurements. I would rather have the CIEXYZ
produced by the raw convertor to check the accuracy but few give
this.It the raw converters let the user access the matrix then that
would be different. If you did deep enough into DCRAW Ccode you will
find the matrixses.
What happens after rendering from CIEXYZ to say sRGb , ProPhoto to
CMYK etc is anyones guess, it up to the user and the control on the
raw converter or what ever software used. All accuracy is then a
matter of choice and gamut constraints.
But as the OP asked about "colormetric accuracy", I think it boils
down to that precise device characterization is essential in order to
produce acceptable scene referred colorimetry. ie device values to
CIEXYZ with minimum errors.
Iain
_______________________________________________
Do not post admin requests to the list. They will be ignored.
Colorsync-users mailing list (email@hidden)
Help/Unsubscribe/Update your Subscription:
This email sent to email@hidden