Re: digital negatives, just a thought

Dave,

you are right to a certain extent about working with as many bits as
possible and doing all transformations at this level before going back down
to 8 bits, you loose much less this way, lets say it is a minor loss and we
are both happy.

A couple of things we can't avoid at all though, the printers output is not
linear with respect to density and of all the values it can produce only a
subset is effectively used for the print emulsion. There is also the
emultion response itself with its own requirements which puts on further
constraints. I think we could call these combine effects granularity or
something similar.

Maybe we don't need to split hair in four and maybe providing an emultion
with as low as half the potential distinct exposure values we could use is
way more then enough. What I don't get is with the 8 bits / 16 bits thing
where you save only a fraction of an 8 bits pixel value or less then 1 in
absolute terms and I doubt many of you still work at the 8 bits level, why
then trowing half the scale out the window doesn't seem to bother anyone at
all except me and the few others that keep silent, it's insane, that's not a
less then 1 value, it's at least more then a 100 values that become useless
with most process. Imagine the delicacy with which you could tailor the
response of your emultion if you could use the full 256 values your original
is made off (at the time we feed it to the printer). With a wet negative you
can't modify the shape of the response curve, it's build in, but with
numbers that's another story.

One last try:

............................................................................
............................................................................
.................................................................

or

****************************************************************************
*********************************

That's exactly how I see it and how it is in practice, we have no choice, as
soon as one tries to get the image out of this virtual world, there are
plenty of physical constraint we have to deal with and instead of trying to
fight them as we do with curves, another way is make them work for us. I'm
begining to have second thought about all this, I'll put a team on it and
see what comes out...

Regards
Yves

----- Original Message -----
From: "Dave Soemarko" <fotodave@dsoemarko.us>
To: <alt-photo-process-l@usask.ca>
Sent: Thursday, May 04, 2006 11:40 PM
Subject: RE: digital negatives, just a thought

> Yves,
>
> It is true that when one can utilize all the 256 colors, a pretty smooth
> gradation can be obtained. It is also true that when you have a 8-bit
> (256-tone) image, and you apply some curve, your image degrades (several
> tones can merge into single tone and there is tone jump).
>
> But that is why most people would capture the image in higher bit depth
> (like 12 bit), do the adjustment there (and it is be done correctly), and
> then finally transform it to 256 tones (8-bit) before printing.
>
>
> Dave S
>
>
> -----Original Message-----
> From: Yves Gauvreau [mailto:gauvreau-yves@sympatico.ca]
> Sent: Thursday, May 04, 2006 12:18 PM
> To: alt-photo-process-l@usask.ca
> Subject: digital negatives, just a thought
>
> Hi,
>
> Lately, as you may be aware I have said and wrote many things about
digital
> negatives and a new concept I came up with but now this message is about
> putting all this in persective and explain in as simple terms as I can
what
> I mean by all this.
>
> The premise of all this is, information lost, you may not be aware that
most
> printer, at least those I have used, can only output 256 tones in B&W and
16
> million plus colors which comes from the combinaison of 256 values of red,
> 256 values of green and 256 values of blue or 24 bits if you prefer. I
think
> most of you may print digi-negs in color but you use none the less only
256
> colors what's worst is when you use a curve and here is why.
>
> I'll use only the 256 B&W tones in order to keep things simple but the
same
> can be said about color. Mark told us the other day that it was possible
to
> obtain a (UV) density as high as 4.0 from a printer and most process need
> less then half this value to do the job right. What happens to the other
> half you don't need? Well it is lost, another way to say this, is out of
the
> 256 tones or densities a printer can produce the process requires only a
> (small) subset of that. The less exposure range a process needs the worst
> this information lost becomes. In other words you are putting on your
> digi-neg just a few distinct densities and it is bound to show on your
> prints in some way.
>
> The concept I'm working on would use all the 256 tones present on your
> original and translate them into 256 distinct densities that spread the
> exposure range required by the process you intend to use and that's a
world
> of difference. Like I said to Don Bryant, it would be like choosing
between
> a bicycle ride and a F-18 ride. The concept resemble what we do when we
> create a profile for a printer, when we do this we are interested in how
our
> printer spits ink on the medium from the various input values basically
and
> we adjust what we feed the printer such that the output meet our
> expectations color wise, all this is done behind the scene with color
> managment.
>
> With this new idea of mine, we would do something very similar, we want to
> know what UV densities each of the 2^24 possible colors can produce.
Knowing
> all this we could choose only 256 from of all those colors that spreads
> uniformaly the hole range of densities required for a particular process.
In
> other words absolutely no information loss, better yet do you think out of
> these 16 million colors of which a printer can only print a subset but
still
> a very large one, could there be enough distinct densities that we could
> translate a 16 bit original B&W into 65k uniformly spread densities from 0
> to whatever a particular process requires? I don't know yet but I just
know
> any number larger then 256 would be an impressive improvement.
>
> I ear already voices saying how will you do this, is it practical to
measure
> 16 million colors, is it feasable, will it be easy to use, do we need this
> level of control, etc. At this time I would be more then satisfied to
reach
> a one to one translation for an 8 bit original B&W or 256 tones to 256
> densities if you prefer. I hope everyone knows that a specific value of
> density on a negative translate to a specific value of exposure for the
> print, just in case and the more distinct densities we can use the more
> continuous the tonalies on the print will appear, think of what happen
when
> you use a step tablet you get a very visible stair case effect in the
> tonality gradation. Increasing the number of step to an as large number as
> possible will simply make the steps practically invisible to the naked
eye,
> it is as simple as that.
>
> I don't think it would be reasonable to expect that someone would measure
16
> million possibilities manually but a computer could. Ultimately what we
want
> to know is how a particular process will translate the input we give it in
> the for of a negative and that's easy to find out. I could refine this
later
> but for now let just say that I create an 8x10 color image with about
50000
> 1mm square each having a distinct color, I print this on a transparency
and
> I use it to expose a particular emultion. I would only need to scan this
> print, look at the histogram and I would know in a second if I have a full
> range of tones from 0 to 255 but it is most probable that I don't actually
> need the values to spreads from 0 to 255 because the paper is not
perfectly
> white so I shouldn't get as high as 255 values and the Dmax of the
emultion
> may very well not translate to a 0 black. What I'm looking for then is
that
> the histogram as no empty bin sort of speak from the minimum to the
maximum
> values I can read back from the print and the best way to know this is to
> make a 16 bit scan and have a look at the data directly. Unfortunately,
> photoshop and other similar programs won't be of much help for this but
this
> may not be such an handicap, someone could write a little program to do
just
> that and we would need one to translate our 8 bit original into the the
> proper colors. I'm sure this is kid stuf.
>
> I have other and possibly better ideas to find out the net output of a lot
> of colors would have on a alt-process print but for now lets keep it as
> simple as possible. Just like profiling a printer with the right tools is
> relatively easy, this concept would require about the same amount of work
as
> for a printer profile, print a large number colors on a transparency, use
it
> to make an print, scan it back, feed this to a program that generate a
> mapping for each 256 values an original can have, then every time you want
> to print something using the same process and parameters, feed this to the
> translater program with the appropriate "profile" or mapping et voila!
Just
> like you would do for each printer, paper and inks combinaison you
profiled.
>
> If anything about this concept is still unclear to you do not hesitate to
> ask me directly or through the list. To resume all this, I think the step
> tablet idea is the easiest to understand, if there is not enough usable
> steps on the step tablet or ultimately on the negative the results can be
> far from an apparent continuous gradation of tonalities from the lightest
to
> the darkest tones of the print, as shown quite visibly with just the few
> step available on step tablet.
>
> Regards
> Yves
>
>
>
>
Received on 05/05/06-09:25:57 AM Z