Digital BW, The Print

Hi Tyler

> 
> I don't think they've ever been the same nor should be. BPC is not
> "built in" to perceptual rendering.

The Perceptual intent requires that black points be mapped.  So "black point
compensation", in the general sense (as opposed to Adobe's specific
algorithm), is required.  There are, however, no specifications as to how
this must be achieved.  In fact, as you note below, there are no rules as to
how the Perceptual intent is to be managed.  It is entirely up to the
profile maker (or software provider) - so-called "secret sauce".

> Something similar to it is an
> natural consequence of remapping/compression to the destination gamut.
> RelCol with BPC, in theory and in my experience, is supposed to leave
> much of the scale alone, and simply begin to compress at some point in
> the shadows eventually reaching an in gamut black point at the end.
> You will see middle gray shift much more in a perceptual conversion
> than a RelCol with BPC conversion.

We are really talking about the same thing.  A straight Relcol has no black
point mapping so Adobe introduced BPC to plug the gap.  Perceptual requires
black point mapping but how you do it is not specified.  In a colour world
there are very real differences between these two intents because of how
non-greyscale out-of-gamut colours are managed (ie shrink the entire gamut
to fit, perceptual, or render the next best in-gamut colour, relcol).  In a
greyscale world, though, there aren't any out-of-gamut colours remaining
once you have mapped the black and white points.  So if you were approaching
the issue purely from a greyscale world perspective then you'd be facing an
identical technical issue.

The bit I don't get is this.  The greyscale axis is a straight line up
through the middle of a gamut plot.  If I had to create a transform that
mapped both black and white points be it either in one go like perceptual or
in two steps like relcol with bpc then I would expect the same transformed
greyscale in the end.  Obviously this is not the case.  The fact that Epson
prefers its own "black point mapped relcol" transform could suggest,
perhaps, that they are not happy with Adobe BPC.

At the end of the day, the ICC spec has dictated how white points should be
mapped but not yet dealt with mapping black point (it's not part of relcol
and perceptual methods are not prescribed).  So it's up to the software
provider to think of an algorithm that makes the most sense. I guess I
wouldn't have thought that there would be too many
scientifically-supportable alternatives but maybe there are...

Steve







>If I recall from months ago,
> maintaining the middle gray value in your conversions was desirable to
> you, therefore perceptual may not be useful.

I would say my thinking on this has changed in that I prefer a more
perceptually pleasing change in tone from black to white.  What I have found
though thus far is that the QTR Create ICC approach slots that mid grey
right about where it is in the file workspace.  For colour work, I much
prefer relcol over perceptual.

> You will also see, strangely, with CMYK profiles, a DIFFERENT black
> value (CMYK ink combo) resulting from the two different kinds of
> conversions, with some profiles.
> Of course, since profile builders can make perceptual tables however
> they like, and BPC is not an icc standard, but an Adobe implemetation,
> many of these issues are not cast in stone.
> Tyler
>