Digital BW, The Print

Hi Martin,

Thanks for replying, I was feeling a little like my post went into a black-hole :).

I think you bring up some very interesting comments about tones, number
of tones and tonality.  Images in the "real world" or "analog world" are 
completely continuous in all 3-dimensions of the Image Model that I talked 
about in the last post. Let me make it clear what I mean by some examples.  
First the spacial dimensions, all points (and there are an infinite
number) between the left boundary and the right boundary exist.
There are no holes.  If you looked at it as a number line, its a solid line;
its not a series of discrete points.  You can't come up with a point in space
between left and right that is illegal or not part of the image. Likewise,
the tonal dimension is also continuous.  All tones are possible.  If you were
to look at a smooth gradation in tones of a sky all tones from the darkest sky 
tone to the lightest sky tone are represented.  If you were to graph light 
intensity you'd get a nice continuous curves -- no discontinuities.

Aside for Quantum Physics affectionatos:
I'm well aware of quantum physics at the nuclear level.
This is way, way, way out of the realm of human perception and
certainly has no place here.  

Now when we get to all of our Imaging Systems lots of things happen.
The real world continuous Image has an infinite amount of information.
Since all three dimension are continuous the some total is infinite.
The saving grace is that human perception, i.e. sight, can in no way 
gather or process and infinite amount of info.  The limiting factor
is the "resolution" of our eyes.  We have discrete rods and cones in
our eyes and relatively poor optics in the lens of our eye.  Fortunately,
our brains have an amazing capability to integrate "light" info
over both time and space amassing much more "sight information" than
the hardware (the eye) seems capable of.

Anyway, precisely because of the human perceptual limitation we
can gear and design the Image Systems to be only as good as necessary.
We only need the output to be sufficient to mimic the perception
of the original image.  The curious thing here is that the reproduced
image can have drastically different information than the original.
The important thing is ONLY that it stimulates the eye and brain
such that the "perception" is as close as possible.

I'd like to illustrate this important point with a short aside to
the workings of color reproduction.  As most of us know visible
light in the real world is a small, continuous, range of 
electromagnetical radiation.  We call it the spectrum from red to
violet.  In a rainbow humans can see this whole spectrum.  Now 
display that whole spectrum on a computer monitor.  We "see" the
same whole spectrum (at least within the gamut of the monitor).
The BIG, BIG difference is that the monitor only puts out three
distinct frequencies of light.  That's why call them RGB (Red,
Green, Blue).  The only thing magic or special about R,G,and B
is that those colors match the sensitivity of the three types
of cone receptors in the eye.  Our perception of color is by the
brain comparing the balance of light intensities of each of the
3 types.  The RGB monitor need only stimulate the 3 cone types in
the correct balance to produce the perception of any color.  The
point is all color imaging that we do: from film to color printers
the entire 3 color system is designed based on the human eye,
not at all trying to reproduce "physics" notion of the color
spectrum.

My major point here is that: Imaging is always based on trying to
reproduce the human perception not trying to reproduce reality.

Back to black and white Images and Imaging.  As I said earlier,
the human eye has quite a few limitations in resolution and optics.
However what we perceive via the processing of the brain is
significantly better.  Integrating over time (we can watch movies)
and space (the eyes are almost constantly moving -- the receptors
pass thru all points in space) our perception is quite continuous.
Our challenge in Imaging is to provide just enough data so that
the brain can perceive a continuous toned image.  In traditional
photography film is an analog medium based on grains of silver.
The grains themselves are discrete, but given sufficient fine
grain and limited enlargement we can perceive continuous tonal
gradations.  In fact even with very coarse grain, enlarged a lot
we see the grain but we can still "feel" a continuous gradation.
I recently saw a Salgado exhibit that had prints up to about
20x30 inches from high speed 35mm film.  The grain was huge but
stand back a few feet and the tonality was beautiful.  Certainly
not lack of tones.

Digital Imaging presents a whole new set of issues.  First the
basic notion of digital is quantizing values, both spacially
and tonally.  We now have the situation of "holes" in our nice
continuous images i.e. discrete information in discrete places.
This may seem inherently a no win situation, but remember all
that really matters is reproducing a "perception" of a continuous
image.  Obviously, how close all that discrete digital data is
determines whether we get a continuous perception.

There's an awful lot of talk and effort about distinguishing
gray tones.  Given our universally common 256 tones of gray, the
effort always seems to be to figure out how to distinguish as 
many as possible.  While this is in general a noble and worthwhile
effort, there's another side of the story that I think is worth
thinking about.  Its good to have distinguishable gray tones but
it is also very important to have indistinguishable tones. By this
I mean there should be tones that are different in an absolute
sense -- a high quality densitometer would see the difference --
but they are humanly indistinguishable.  Why you say?  Well 
consider a very gradual sky gradation, if all the grays in that
gradient are distinguishable we will perceive a step function in
the tones.  Say the tones are grays 140, 141, 142, 143, 144.  We
want any two adjacent tones to be indistinguishable but any tones
separated by 2 or more to be distinguishable. This will produce the
largest number of distinguishable grays and still allow a continuous 
perception. This is probably one motivation for going to 16 bit tonal
representation -- we want to make sure there are indistinguishable
grays. It sure sounds ironic, but with 256 new gray values between
each original gray you're guaranteed to get some indistinguishable!

The other major factor in producing the perception of continuity
is Noise.  I think in general noise is thought of as a detrimental
factor -- we strive to eliminate it or drive it to zero.  In fact
noise is a wonderfully useful tool to improve the perception of
tonal continuity.  Recently someone in this group noticed that when
using a very simple "gradient" tool in Photoshop, they didn't get
a perfectly flat and pure histogram.  You might think it ought to
get a perfect gradient from dark to light, but in fact the folks
at Adobe, in their extensive knowledge of Imaging, know that the
pure gradient actually doesn't look "good".  In variably, there
distinguishable tones adjacent and the result is perceiving bars
in the gradient.  The solution -- add noise.  Amazingly, the
gradient becomes very smooth.  Actually, its the perception 
which is now very smooth.  The point is to use the processing in
the brain to the best advantage.  For anyone with a sky gradient
which is too posterized adding noise can be a tool to help!

Well, that's about it for now.

Regards,
Roy


--- In DigitalBlackandWhiteThePrint@y..., "Martin Wesley" <mwesley250@e...> wrote:
> Roy,
> 
> Thank you for your well thought out post on Imaging. I refrained from
> responding earlier in hopes that someone new would add to your thoughts. As
> you mention at the end of your post people are probably a more than a bit
> reluctant of even mentioning some of these terms at this point.
> 
> I spent some time going through what books I have on photography from a
> technical imaging point of view and the concepts that are of concern were
> much as you have described them. The spatial component of the system is so
> intuitive it is not mentioned except in an artistic manner. What is
> discussed at length is the relationship between the SBR (Scene Brightness
> Range or Scene Luminosity Range), Density Range and Average Gamma (the
> average slope of the line from minimum to maximum which represents
> contrast). From these three they map SBR to the film and then to the paper
> showing characteristic film and paper curves.
> 
> Here in the area of digital printing we still need to define that path from
> mapping film density to the print and digital capture opens up similar
> questions. With the new technology added to these mappings different
> variables may have greater importance at different stages but your 3D
> concept holds a great deal of merit in keeping us focused on what is
> important.
> 
> The idea of distinguishing between tones and the number of tones are
> concepts that I do not ever recall hearing in discussions of traditional
> photography. If not new concepts, they seem to be new to the discussion of
> photography. I believe that they were not discussed in traditional
> photography because they simply were of no interest. The technology of
> traditional photography generally exceeded the ability of our eyes to
> discriminate tones. With the advent of digital imaging, which started out
> with quite poor quality, these ideas became vitally important. Once the
> quality of the equipment, software and materials of digital imaging
> consistently exceeds our eye's ability to discriminate tones the concepts
> will once again hold little interest.
> 
> Thanks,
> Martin Wesley
> 

Roy Harrington
roy@...
Black & White Photography Gallery
http://www.harrington.com