--- In DigitalBlackandWhiteThePrint@yahoogroups.com, CDTobie@... wrote:
>
>
> In a message dated 5/30/07 10:09:17 PM, bob@... writes:
>
>
> > I keep seeing the mention of ICC's with imbedded curves. (I don't
> > think this is what Jonathan meant, IMO.) Is an "ICC with imbedded
> > curves" an ICC plus a curve or a completely self-contained ICC? What
> > is meant by "ICC with imbedded curves"?
> >
> > Thanks, Roy and Tobie, for the very enlightening conversation!
> >
>
> This is the distinction I keep trying to describe, apparently not too
> successfully.
>
> An ICC profile is a three dimensional transform from on space to another, via
> a 3d LUT.
David, I really can't see why you continue to claim grayscale is not an ICC profile.
It's in the specification, it's been there a long time. Anybody can check it out
at: http://www.color.org/
It contains other factors, including a set of single dimensional
> "linearization" graphs for each incoming and outgoing channel. So an RGB printer
> profile comes in from Lab, where you can adjust input L*, or a* or b* values
> before they get to the 3d LUT for transformation. In the case of Create ICC,
> what happens here is the a* and b* values are zeroed out in these tables. Then
> the transform occurs. In the case of CreateICC, this is a dummy transform, a
> simple set of zeros and ones that move the date from one space to the other,
> but don't adjust it. Then there are a set of linearization graphs on the other
> side for the output channels. In a CreateICC profile, it is not the 3d LUT that
> is used for corrections, its the linearization graphs sandwiched to it. Thats
> the "just imbedded curves" that he's referring to.
>
> You could pull this set of curves out of the profile, apply them in
> Photoshop, and not need the profile at all. The reasons (that I'm aware of) for doing
> it in a profile is convenience (you can pick it from a list at print time, and
> don't need to alter your image file) and preview (some CreateICC profiles can
> be used to preview your print results in Photoshop).
No, it's different. Curves in Photoshop just work in the stored values. They have no
notion of the "meaning" of the values i.e Luminosity. Color management deals
with the meaning so it converts the luminosity in the file to the equivalent luminosity
on the print (or screen ofr whatever). Color management always and only deals with ICC
profiles so putting the curves and linearization into an ICC profile opens up the
power of matching luminosity.
>
> On the other hand, a PrintFIX PRO profile uses the 3d LUT transform (a
> different one for each rendering intent, actually) to do its adjustment. If you are
> printing to a "grays only" printing device or driver, thats all you'd get from
> a PrintFIX PRO profile, not because its all it can hold, but because thats
> the only thing being used. So when I build a profile via Canon's "Monochrome
> (Photo)" mode and use the resulting PFP profile to print to that mode, I get only
> linearization out of it, because thats all the mode looks for. No color
> adjustment allowed on the input side. But that doesn't mean the data is "just
> imbedded curves", in this case its "just a 3d LUT conversion" instead.
Using the 3d tables for something with only one dimension is a waste of space.
A typical 3D table may have only 33 grid points per dimension and take probably
a 100 times as much space as a single dimension with a full 256 points. 33 points
connected with straightlines can't be as smooth as 256 individual points.
>
> What I would like people to realize is that they are effectively wasting two
> dimensions of correction (a* and b*) when they use PrintFIX PRO to lineralize
> a black and white mode. If it were physically a black and white printer, sure,
> why not, there's nothing else you could adjust anyways. But if its a highend
> color graphics printer with two levels of gray ink, plus a full range of
> colors, then you have the opportunity to use it all, via color mode, to get the
> most controlled, neutralized results possible, by using all three dimensions of
> the Lab space: L* (the grayscale linearization part that you can get in B&W
> modes as well) and a*/b* corrections, which improve the neutrality of the output.
>
> Sorry to keep at this, but I didn't feel that I had succeeded in expressing
> it in a way that was getting across yet... so I'll say it one more time:
>
> The only method I know of that is designed to automatically measures
> neutrality of adjustable tint output at dozens of levels, and neutralizes all of them
> on the fly is PrintFIX PRO used with the Color Mode of a recent graphics
> inkjet. It produces amazingly neutral output on a wide range of media, with
> controls for tuning how the gray ramp runs to paper white, how much shadow detail you
> get, and other important factors. It produces gallery quality B&W output,
> plus your choice of tints and crosstones, and full color images, on gloss and
> matte media, all from the same printer; all from the same paper profile for that
> matter (assuming the printer has two levels of gray ink, and both types of
> black ink). Given all of that, I fail to see why people that already have
> PrintFIX PRO at hand still want to work through the B&W mode of such printers, and
> live with whatever wiggles and wabbles the neutrality of that mode produces on
> the paper they are using.
People want all kinds of things. For years they've spent lots and lots of money
on custom RIPs and drivers so they could control what gets on the paper.
I just did a test of ink usage in the Epson driver. The fact that the regular driver
uses quite a bit more yellow ink may be a good enough reason for some to want
to use the ABW driver or a custom RIP.
I think this whole thing has been beat to death.
Roy
Show quoted textHide quoted text
>
> C. David Tobie
> Product Technology Manager
> ColorVision Business Unit
> Datacolor Inc.
> CDTobie@...
> www.colorvision.com
>