Digital BW, The Print

I posted this a while ago, and it seems yahoo lost it.
If it turns out as a duplicate, sorry.  --Roy

--- In DigitalBlackandWhiteThePrint@y..., "Martin Wesley" <mwesley250@e...> wrote:
> Roy,
> 
> A good post and I concur, for what it is worth. I am sorry I dropped out of
> the discussion but it did not seem to be leading anywhere constructive.

Thanks, Martin.  
> 
> I favor the simple explanation when it will do and the classic view of the
> dynamic range in photography as being the ratio of the useable/meaningful
> max and min seems to me to suffice quite nicely.

Simplicity in concepts is so important that it must be emphasized over
and over.  It has to be simple in the math, simple in the physical phenomena,
and simple in the human perception.  There's a philosophy caller "Occam's Razor".
For anyone not familiar with it, this dates back centuries to the 1300's to
a philosopher, William of Occam.  If there's any philosophy that I'd call
universally applicable to life, this is the one.  Go to www.google.com and search for
occam's razor --  its truly worth reading for any level of person for any
walk of life.

> 
> The one thing that I will add is that in both the audio and the imagining
> models there is no mention of discrete or finite numbers of volume levels or
> brightness levels. The implied assumption is that any level in between can
> be used or achieved. Even if theory suggest a finite number of levels
> actually exists, this seems totally irrelevant due to the factor of human
> perception. If we could clearly hear or see distinct level changes in the
> signal we would simply reject the audio or the image as being of
> unacceptable quality.

Yes, the physical phenomena of both light and sound are analog i.e. completely
continuous.  Similarly the human perception of both is also completely
continuous.  There are bounds (max and min) but between those its
continuous.   Discrete levels of either sound or light are only introduced by
manmade devices -- usually digital processing.

> 
> Thanks,
> Martin
> 

I had one email question that is relevant and I think its worth addressing here.

What about noise?  Where did that go?

Let's look at the audio realm and how it relates to the DynRange formula.

DynRange = 10 * log10 (maximum signal power / minimum discernable signal power)

The word discernable is where the noise factor comes in.  In audio the idea
is that at the low power end, the signal has to be strong enough to be
detected ABOVE the noise level.  Signals that are below this level cannot be
perceived or measured and should NOT be considered as part of the DynRange.
Noise is THE main limiting factor for the lower bounds of "useful" signal.

So how does this relate to imaging?
First I think there is a semantics or language problem.  We use "noise" a whole lot
in sound:  everybody has had the experience of sitting is a "noisy"
restaurant or bar.  People can easily relate to the word and idea that in the noisy
bar, you have to shout for the other person to hear you (sound familar?, the signal
has to be greater that the noise in order to be discernable).

I've never heard anyone look at a scene in the real world and remark:  what is
see sure is noisy.  The word from an language point of view doesn't have an
obvious, customery usage in imaging.  But, of course we're all somewhat savvy
to technical terms and we have a feeling of what noise is.  I think about the
only way to talk about noise in light is from the point of view of something
detecting or measuring light levels.  In measuring a light level, if the image light
level is so low that extraneous light masks the image, then the image light
it down below the light noise and therefore undiscernable.

Lets try an example.  Scene of a white barn out in the sun, there's a door open
that goes way back to a distant wall.  There are no windows so the inside
is very dark.  Lets measure light using our eye and using a spotmeter.  From
the outside neither our eye nor the meter can get a "meaningful" reading.
We can go inside and both the eye and meter work perfectly, they are both
plenty sensitive: we can see and measure lots of different light levels
inside the barn.  Back outside again, we see a black hole: nothing whatsoever,
the meter comes up with a reading but its entirely governed by flare and
other extraneous info.  This is what I would call noise in the scene.  There are just
too many extraneous light levels to detect the real image light level.  For
the eye case I'm sure there are light waves hitting the retina is the area
corresponding to the dark door.  But the brain is well aware that the noise
in that area is dominating and so doesn't bother us with a useless perception.

This idea of noise is what I think is relevant to Dynamic Range.   Its just
a matter of image signal must be greater than the extraneous light
energy for the image signal to be "useful"  and therefore within the DynRange.
This concept is pretty easy to see in real scenes but when talking about
image prints most of this isn't very relevant.  A print on paper is an abstraction
of a real world scene.  The real world scene can potentially extend without
bounds towards dark and towards light.  There's no inherent limit to the
max or min energy in the light waves.  On a print the paper has absolute
limits the white of the paper and the black of the most black stuff we can
put on the paper.  In an appropriate light, we can easily see or detect all light
levels that could possibly be emitted.  So the issue of noise (extraneous
light) limiting the max or min light level from a print just doesn't exist at all.
(Aside: viewing a print under glass in a glary area would be an example
of an "inappropriate" light that would probably limit our seeing max to min).

Before there's a bunch of objections, I think there is also a very different
phenomena that a lot of people might think of as noise in imaging.  Many
imaging technologies represent light values in discrete chunks.  Film is all
based of grains of silver, prints are grains or for inkjet they are dithered
dot patterns.   They are all based on representing continuous light data
as many little pieces that all averaged together to make our perception
appear continuous.  The technologies all use various different "chunk"
sizes: some are large like fast film or like a printer with big ink drops;
others are small like slow fine grain film or the newest high dpi printer.
I think we refer to this "chunkiness" as noise as well.  The noisiness is
function of how much stuff we perceptually have to average over to make
our perception appear continuous.  If averaging is too small an area we see the
chunkiness and call it noisy, if the averaging area is larger the chunkiness
goes away and its less noisy.   Note that this phenomena is just in the
imaging technologies, there's nothing is real scenes that corresponds
to this.   This kind of noise is certainly has has a long history in imaging
and is important as a concept in imaging, but I don't think it plays any role
in the Dynamic Range concept.

Here's an illustration:  Imagine a print made from a very grainy piece of
film.  The print spans the whole paper range -- bright white area to
deep deep blacks -- and of course there's lots of grays in between.
Look at the print from close up.  There's lots of grain, lots of chunkiness,
and lots of noise.  Now back up, our eyes now have to average light
info over a larger area due to limited eye resolution.   Its exactly the
same print but the grain, the chunkiness, and the noise all start to
disappear.  Soon there is just completely continuous perception. 
The max white and max black are still the same light levels.
Has the dynamic range of the print mysteriously changed?  Shall we
talk about a PROPERTY OF THE PRINT which changes even though
its the same print, same light, same eyes?   I say probably not.

Thanks for reading.
-Roy