Re: Is this a scum dot?
Re: Is this a scum dot?
- Subject: Re: Is this a scum dot?
- From: bruce fraser <email@hidden>
- Date: Mon, 18 Nov 2002 10:27:23 -0800
At 11:56 AM -0600 11/18/02, Bruce J. Lindbloom wrote:
Profiles are used to establish the relationship between device values (e.g.
RGB or CMYK) and device-independent values (e.g. Lab). In the process of
creating a profile, we know one side of this relationship (for a selected
set of colors) and we empirically determine the other side. From this, the
entire profile behavior is generalized.
For printer profiles we know the device side (the RGB or CMYK values of the
profiling test pattern patches) and we discover the corresponding
device-independent side by measuring the colors of the printed test pattern.
The ICC specification states that for relative colorimetric rendering
intent, device white (R = G = B = 255, or C = M = Y = K = 0) should map into
perfect PCS white (Lab = 100, 0, 0).
For scanner profiles the opposite is true. We know the device-independent
side (the Lab values of the IT8 patches as found in the reference file) and
we discover the device values (how the scanner "sees" each of these patches,
expressed by its RGB values). The ICC specification does not say anything at
all about the relationship between any device value and device-independent
value for a scanner profile. So when Bruce Fraser wrote:
If the scanner profile translates RGB 255,255,255, relcol, to
something less
than L*100, it is by definition broken.
I'm not sure what definition is being referred to, but I am not aware of any
definition like this. I would instead agree with Graeme Gill that the RGB
value of the whitest neutral patch of the IT8 should map onto PCS white (in
RelCol). One cannot safely make any further generalizations about what other
RGB values must map to Lab 100, 0, 0.
I think the part you're all missing is that you're equating "device
white" with "the white patch on the IT8."
I don't know anyone who actually builds useful scanner profiles this way.
There are basically two rational approaches to building a scanner profile.
1.) Run the scanner "wide-open" and simply characterize its raw
behavior. Don't set end points or adjust the tone curve.
2.) Set the scanner to an arbitrary black point, white point, and
tone curve, then lock the controls.
In the former case, if white on the IT8 produces RGB 255 and black on
the IT8 produces RGB 0, you have a scanner with an extraordinarily
limited dynamic range.
In the latter case, if you set the scanner to produce those values
for white and black respectively, you WILL run into problems when you
try to do real work.
If at some time in the future we have adaptive CMMs that base tone
and gamut mapping on the image rather than the profile, it would make
sense to do something else, but "device white" is by definition RGB
255,255,255. That's the point at which the detector is fully
saturated in all three channels. It can't capture anything brighter
than that, and anything darker than that isn't the whitest white the
scanner can produce. If you map a lower RGB value in the profile via
relcol to L*100, you're throwing away all the scanner's response to
brighter values -- they all get clipped to L*100.
If you map "reference white" (the brightest white on the target) to
"device white" you're making the assumption (which experience tells
me is unwarranted) that the "reference white" is the whitest white
you'll be asked to scan. That is definitely untrue for
batch-manufactured IT8 targets.
In practice, RGB 255 is not a value you ever want to capture in a raw scan...
--
email@hidden
_______________________________________________
colorsync-users mailing list | email@hidden
Help/Unsubscribe/Archives:
http://www.lists.apple.com/mailman/listinfo/colorsync-users
Do not post admin requests to the list. They will be ignored.