Hi All,
The ongoing discussion and debate about dynamic ranges and
numbers of tones, etc. seems to just go on and on endlessly
without getting anywhere. All of us in this group are
interested in "Imaging" in some form and at some level. I've
written the following with the idea defining Imaging, seeing
where the ideas come from, how they relate, how we talk
and communicate amongst ourselves and others about Imaging.
Its a little long winded but I felt I need to get my thoughts
down.
IMAGING
What is Imaging? I'm talking about anything and everything
we SEE with our eyes. Thus X-Ray Imaging, Sonic Imaging are
all out of the discussion. Imaging is really DEFINED and
orginates by and because of human sight. Our eyes focus
light from the outside world projecting a 2-dimensional
imaging on the retina. The rods and cones of the retina are
sensitive to light and collect light intensity information
over a two dimensional space. The hardware of the eye in
conjunction with the software of the brain work together to
give us the perception we call "sight".
An Image is a varying intensity of light spread over a
two-dimensional space. Everything in Imaging is related to
the creation, detection, processing, storage or reproduction
of Images i.e. varying light over a two-dimension space.
Certainly, there are many abstractions or intermediate
representations of Images but they all relate back to the
original concept of "varying light over a two-dimensional
space".
Obviously, since "sight" is one of the most basic sense,
the English language (and I have no doubt all languages) have
many words relating to "light" and "varying light". We talk
about light intensity, brightness, dark/light, black/white.
A very key issue is that the light intensity varys; we
are very aware that there is a range of light intensity
that is "useful". We can't stare into the bright sun
because its too intense. Likewise in a totally dark cave
we can't see anything. There is a range of light intensity
that is humanly useful. Light intensity can go above or
below the humanly useful range, but this "useful range" is
what is interesting.
Putting this all together we have a very good model of
what an image is. All told its a 3-dimension object with
two dimensions that are spacial and one dimension that is
light intensity. Another key aspect is that all 3 dimensions
are "bounded" -- nothing extends out to infinity. The
space dimensions are obviously bounded by left and right
bounds in the horizontal direction and bounded by top and
bottom in the vertical direction. The light intensity is
bounded by "darkest" and "lightest".
Thus, an Image is a 3-dimensional bounded object with two
spacial dimensions and one light intensity dimension.
As I said earlier, Imaging is everything is related to
the creation, detection, processing, storage or reproduction
of Images. All Imaging operations take an image (3-dim object)
and at best keep as much information as possible. In
reality most will tighten the boundaries some how chopping
off some of the information.
For a simple example, lets take some photos. We see a nice image
and point the camera and look thru the viewfinder. Immediately
we've trimmed the two spacial dimensions. We now have
much less in view than with our eyes. (Artistically, that's
probably exactly what we want.) Next of course we trim
the light intensity dimension by putting the image on
film. The lightest and darkest we can see with our eyes
is a much wider range than the film in the camera can
handle. The film can only record a much narrower range
of light intensities. It has a smaller "useful" range.
Things aren't so bad because although our eyes can handle
very large ranges of light, often the scene we are
photographing actually has a much narrower range. We
adjust the exposure settings on the camera aligning the
narrower scene range to narrower film range. Later on
when we print the negative we again have the opportunity
to trim the spacial dimensions (cropping) and of course the
more challenging task of taking the range of light intensities
on the film and mapping or transferring them to a piece of
paper. The mapping can involve just shifting darker or lighter,
but also can involve compressing or expanding the range to
better fit the physical characteristics of different media,
In general, all imaging steps put constraints of some kind
on the image information. For an image to go thru many steps
we've got to be careful at every step of the way. The end
result can never be any better than the worst step. Trimming
in the spacial dimensions is so obvious that it probably
needs no discussion, but dealing of the light intensity
data is probably the main effort in all imaging endeavors.
The set of technologies and industries that an image may
pass thru from start to finish is truely amazing. From
light meters, to exposure, to film, to developing, to
photo paper, to another developer, toners, etc. If there's
digital involved there's CCD's, scanners, A/D and D/A
converters, lots and lots of computer software, printer
drivers, printer hardware, paper technology, ink technology.
Get the idea!
Anyway, every single step of the way deals with the range
of light intensity. It's represented in many, many ways,
but there are several universal characteristics. First,
they all have to represent a range of values. They have
to have dark values thru light values. It may be a continuum
of dark to light or it may be discreet values, but there
are always endpoints i.e. max and min. The endpoints
may be fixed obvious points like white paper and paper
with the maximum amount of black ink. Some endpoints are
a little less obvious -- like film. Film itself goes from
fb+f (filmbase+fog) to very, very dense (take film expose
it to lots of light and develop it a lot), but there is
definitely a sweet spot -- its linear between the toe and
the shoulder. This is the useful range for film.
So in moving the image from one step to the next we need
to keep moving the range of light values along the road.
Essentially we want to map the darkest value (and lightest
value) of one step into the darkest/lightest value of the
next step. We are trimming info if that mapping moves
a value outside the next step's useful range and conversely
if we don't use the whole range of the new step we are
using that range to the fullest extent. One very important
characteristic is that given two values (darker and lighter)
they can get closer or farther away tonally. They can even
merge together so they are no longer distinguishable --
lost info of course. But they can never change places.
This guarantees that mapping max to max and min to min
keep the range monotonically intact. We can lose info
due to merging of tones but sometimes that's unavoidable.
What is not guaranteed is that the mappings in the middle
not at all the linear. Its the linearity and the
distinguishability in going from one step to the next that
we spend all our time and effort.
The image with its range of light values moves from step
to step being subjected to the range of possible values
for each step. Each step can represent completely different
technologies. In each technology there are whole sets of
terminology, descriptions and measuring standards for its
dealing with the range of light values. The range of
useful values is called the Dynamic Range. The Dynamic
Range describes the range of useful values, but of course
since the values may be represented in different ways the
units of expressing the range may vary. Some technologies
have very commonly used units and the Range is expressed
in those units. Some use the term just descriptively
to refer to the range but there really isn't a measure
or quoted number for the dynamic range.
Here's a few examples to show some of the variation.
Dynamic Range of film can be expressed as density range
of film itself from the minimum to the maximum useful
density.
Alternatively, it could be quoted in the number of stops
of exposure variation that leads to the above densities.
Dynamic Range of a scanner is typically quoted as the
maximum density of the film that the scanner can get
detail from. Here, minimum density is assumed to be
zero, so it is a range.
For a final print Dynamic Range is just the white of
the paper to the max black ink we can get.
A particularly curious case is the data file inside the
computer. If its a 8-bit grayscale, by convention 0 is
black and 255 is white, if it was a bitmap (1-bit) there's
just 0 and 1. In a sense they both have the same
dynamic range but being unitless and totally arbitrary
it probably not very informative.
The point here it that the term "Dynamic Range" is used
in many different contexts both as a calculatable measure
and as a descriptive label for the range. The key is that
all the contexts are tied together by the mapping of
the range of one step onto the range of the next step.
And the ranges are all ranges of light values from
dark to light.
--------------------------------------
Obviously, there are other viewpoints about using this
term "Dynamic Range" (Hi Austin). But I've looked at quite
a few specs and discussions using this term, each and
every one in Imaging uses it in this context. The range
of light values from dark to light is unquestionably the
most basic property of sight. Austin's Range of Contrasts
from min contrast to max contrast, leaves me and I believe
others with a "What the hell is it?" feeling. I've
seen several people innocently ask about Dynamic Range
in a context (I believe) much like my description and barrage
of discussion that gets nowhere ensues.
The regulars here now avoid saying the term like they
avoid the plague.
For me it boils down to two issues:
1) Range of Contrasts doesn't help my understanding of
Imaging i.e. it doesn't go with a perceptual property
2) Nobody else in Imaging uses the term that way.
---------------------------------------
Term usage aside, I found the exercise of thinking thru
Imaging to be helpful to myself. Comments?
Hopefully someone else will find this useful.
Roy
--
Roy Harrington
roy@...
Black & White Photography Gallery
http://www.harrington.comMessage
Thoughts about Imaging
2002-04-01 by royvharrington