Re: TECH-3D digest 176

[T3D John Ohrt <johrt@xxxxxxx>]

T3D Eric Goldstein wrote:

> > Monitors have a much greater dynamic range. (than paper prints)
> 
> I'm told this is true and my eyes concur, but can anyone give be a number?


Yes and no.  One of the great differences between film and electronic
devices is the transfer function.  In film, we can regard it as a
relationship beween photons and density assuming no reciprocity
failure.  In electronic devices, you can regard it as photons and
electrons.  The problem is that the film characteristic doesn't have the
same general shape as the electronic devices, and varies widely.

The transfer function for electronic devices is often very closely
matched by a gamma function.  By applying the inverse gamma function,
the data can be made to have a gamma of 1.0 (ie. the data is linear). 
This is very important for digital image processing as any convolution
or deconvolution process will not yield the expected results unless the
gamma of the data is 1.0!!!!!!  eg. any filtering operation is a
convolution operation!!!!


Good news:

Mac users have display subsystems with a gamma of 1.0 (nominal) which
makes wysiwyg straight forward.

Unix users can have a display subsytem of any gamma, if your hardware
supports analog gamma correction, gamma 1.0 data looks great. If the
correction is done digitally, but transparent to the data, it looks ok
and IMO, is good enough for a lot of work.

Bad news:

MSDOS and OS/2 users have gamma=2.2 display subsystems, so your photo
processing software needs to provide the transparent conversion for
you.  This problem is now fairly well recognized and there is a why to
accomodate gamma=1.0 data in most systems via the photoprocessing
software similar to the digital correction available under Unix.

WARNING WARNING WARNING - if your image processing software does not
support floating point data and files, gamma correction really chews
into the image quality.  If the data avoids filtering etc., then do the
gamma correction LAST.  If convolve and deconvole are essential, try
them with and without gamma correction before the operation.  Often the
results of nonlinear data look better that the results after one or two
gamma corrections.

NOW HERE IS THE BIG PROBLEM -  We are about to compare apples and
oranges.  Film density does not take into account linearity and
electronic data does!

So, with all that in mind, the density range of an electronic system
whose gamma=1.0 is the log of the number of bits per channel. log256 =
2.4

By comparison, the density of an entry level digital printer may not
exceed log32 = 1.5

Conversly, a state of the art monitor can display log4096 = 3.6

A state of the art CCD camera is limited in accurracy and dynamic range
by the ADC and the time permissable for conversion.  Exceeding 16 bit
accuracy and dynamic range is doable for big bucks log65536 = 4.8

While you can't display such data in one shot so that the eye can
resolve all of the detail, scientific viewers allow for you to hunt for
targets within the data.  Some of these viewers are freeware.

Please remember these are truly linear representations, no ifs, ands,
buts or maybes.

Please people, don't let this evolve into a pro and con of analog vs
digital processing.  Just use whatever works for you.

Please note that with high numbers of bits per channel, multiple scans
with different settings etc. you can scan a film so the output data is
highly linearized.  The question is, do you want to?

Some of the most stiking photographic art uses the nonliear transfer
function to create the artistic effect.  "Correcting" such data would be
akin to colorizing all black and white photos!  But scientific photos,
then it may be desireable.

It is notable that most computer quality image hardcopy is accomplished
through photographic means or hybrids thereof.  No one's digital printer
can beat the color rendition of a good 35 mm photo of a monitor.  Image
setters are currently the best output device, up to 4000 pixels per inch
with 8 bit channel accuracy are available in larger cities for 35mm
slides. (they still haven't caught up to scanners)

Hope that helps, but I fear I have muddied the waters.
--
John Ohrt,  Regina, SK, Canada
johrt@xxxxxxx


------------------------------