Re: Epson canned profiles
Re: Epson canned profiles
- Subject: Re: Epson canned profiles
- From: ben <email@hidden>
- Date: Tue, 17 Apr 2018 09:15:38 -0700
- Sun-java-system-smtp-warning: Lines longer than SMTP allows found and truncated.
On Apr 17, 2018, at 6:02 AM, email@hidden wrote:
> If you inhabit the "dE differences of .85 at the 6ish L value level make no
> difference”-world then it does not discount all the people who that would
> make a huge difference to.
I've avoided getting into _why_ Andrew is spouting nonsense about the supposed
insignificance of 0.85 dE because I couldn't even get him to put any actual
data to his claims, but I think it's very much worthwhile having a side
discussion on that point.
Yes, Put a pair of patches side-by-side that differ by 0.85 dE and most people
would be very, very hard pressed to tell you which is which. Separate the
patches by even a small amount of space and zero humans will do better than a
coin flip to tell you which is darker / more saturated / whatever.
...but that's not the only context in which humans are exposed to relative
color difference. It's not even remotely the most significant context,
especially with respect to profile building.
First is that we're especially sensitive to changes in hue near the neutral
axis. Make that 0.85 dE a comparison between L*a*b* 50, 0, -0.43 and 50, 0,
+0.43, and a *LOT* more people will be able to tell you that the one patch is a
bit warmer and the other a bit cooler. But a 0.85 dE between two points on the
spectrum locus and low brightness? Toss a coin.
But even more important, especially with respect to profiling, is that you can
pick basically any pair of colors that differ by 0.85 dE, place them adjacent
to each other with no space separation, and the line dividing them will be
instantly visible to everybody. You might not be able to describe the
difference, but that there's a line between them will be obvious.
Which is why we don't use seven-bit color.
I would hope that everybody here is well familiar with how easy it is to see
banding in eight-bit gradients. Banding in a seven-bit gradient is far more
glaring -- and the difference between steps in a seven-bit gradient is a mere
0.78 dE.
So Andrew's blather about how a difference of 0.85 dE in profile building is
invisibly meaningless is exactly equivalent to a claim that 7-bit imaging ought
to be good enough for anybody, and his pouting that you can't identify the
difference between 0.85 dE samples is irrelevant.
It further illustrates a very disturbing shortsightedness on Andrew's part.
Profiles can be used for spot color matching, yes, but that seems to be the
Alpha and Omega of Andrew's goal in profile building. For everybody who's not
printing corporate brochures where the company logo has to be the right Pantone
number, profiling is all about image quality, especially fine detail. Open
shadows and clear highlights in black-and-white fine art; smooth (and natural)
skin tones in fashion and portraiture; that sort of thing.
Missing the black point by 0.85 dE might not be as big a deal for a printer as
well linearized as the new Epsons appear to be, but that would be entirely
thanks to Epson engineers. Indeed, I wouldn't at all be surprised if you'd get
more detail from a matrix profile than from one made the way Andrew's been
describing.
...which is why context is so important. Miss some super-saturated yellow by
0.85 dE and nobody will care. Miss the black point by 0.85 dE and some will.
Miss the black point by 0.85 dE in one direction, the midpoint by 0.85 dE in
another, and the white point by another 0.85 in yet another direction, and
you've got a serious problem.
Then again, Andrew has repeatedly repudiated his own data, so who knows what
he's actually doing?
Cheers,
b&
_______________________________________________
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