Re: BPC
Re: BPC
- Subject: Re: BPC
- From: Klaus Karcher <email@hidden>
- Date: Tue, 23 Oct 2007 03:35:47 +0200
eugene appert wrote:
nothing about how it translates values above
the black point or about how it is affected by the a* and b* axes
when compressing the L* axis.
see 7.3.: Computing the mapping from SourceBlackPoint to
DestinationBlackPoint.
BPC is a uniform scaling in XYZ space. The center of the scaling is the
white point of the output profile, that means every color will be moved
radially towards or away from it. The larger the distance to white, the
stronger the change. Colors close to the neutral axis will basically
change in lightness whereas light, chromatic colors like yellow will
mainly change in cromaticity.
Example: 70% Yellow in ISOcoated_v2_eci.icc
will be mapped to
CMYK 2 0 94 0 (Lab 79.7 -2.5 58.6) rel without BPC and to
CMYK 2 1 67 0 (Lab 79.9 -2.0 49.0) rel with BPC
in ISOnewspaper26v4.icc
Ths example proves that BPC is -- other than often claimed -- no a
directional scaling along the L*-axis, but rather a simple way of gamut
mapping. The result is satisfying as long as the gamuts are similar in
shape (often more satisfying than perceptual in this case because it
hardly causes distortion), but it can induce serious clipping if the
gamuts differ in shape.
Apparently BPC is just as blind to file data.
It is absolutely blind to image data. The "smartest" thing it can do is
to calculate the black points of the profiles involved in a conversion
at link time (these points were already calculated at the time the
profiles were build: they are needed to map the device black point to
the PCS in the perceptual tables).
As there is no substantial difference in black ponint mapping between
BPC and the perceptual intent, one would expect similar black points
regardless of the transform (perceptual without BPC, percerptual with
BPC or relative colorimetric with BPC), but in practice the differences
can be considerable
Example: if you convert CMYK 83 79 73 95 from ISOcoated_v2_eci.icc to
ISOnewspaper26v4.icc, you will get
CMYK 58 45 40 94 perc without BPC,
CMYK 54 40 36 89 perc with BPC and
CMYK 50 38 36 90 rel with BPC
I have noticed an anomaly, which occurs when printing a colour image
with a large area of subtle gradation from deep shadow (L*1 to L*
20). Soft proofing this image using BPC produces a horrible
posterizing or banding which is often even worse when actually
printed.
I am curious as to why BPC fails
I think what you observed is the result of a bumpy, not very
well-defined PCS<->device relation at the bottom of the device gamut.
Problems like this appear frequently and can be caused by measurement
errors, inappropriate separation parameters, bad device calibration,
wrong ink limitations (maybe enforced by the printer driver or RIP),
profile interpolation errors or simply strange device behavior.
Probably one of these reasons (or a combination of several of them)
triggers the mismatch in estimating the destination black point. Maybe
you can avoid it simply by using the perceptual intent. Otherwise you
have to revise the above-named issues and re-profile your device. If you
are still not satisfied with the results, use device link profiles.
Klaus
_______________________________________________
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
| References: | |
| | >Re: BPC (From: "eugene appert" <email@hidden>) |