dealing with color negative films

[Dale Kelly]

color negative films are designed to make good prints, either on
photographic paper, or motion picture print film

- there is no "scan only" color negative film designed for accurate scene
reproduction
- there is no "film recorder" color negative film designed to work
specifically for a film recorder

while the above are ideals considered in the design of the film, the main
consideration is a good print, and thus the ICC specification for scanning
a color negative film is the characteristic intent of the film when it is
printed

to model a color negative film you must keep in mind there is considerable
processor and printer variability and that the center population of the
film depends on the process and printer used in the design of the film, in
addition, the color negative system as a whole is multi-stage and
considering these things it does not lend itself to a target/fingerprint
characterization

I worked for kodak for 10 years and the following is how they deal with
scanning and recording films in-house:

to make a profile for a color negative scanner:

1) set the scanner to linear transmission or log density if it has a log
amp, any other setting requires unbuilding and introduces encoding noise,
the scan can be consider channel independent exposure if the film is a
perfect scene capture material, which it is not

2) process the transmittance of the film through channel independent
sensitometry lookup table with quadratic interpolation, which models the
contrast of the chemical process, this gives you channel independent ddye
transmission of the negative's dyes

3) process the channel independent dye transmissions of step (2) through an
interimage matrix which models the various layer to layer crosstalk of the
chemical process, there are various considerations of dye wandering and
dye scavenging because of the various dye characteristics or DIR, DIAR dye
couplers introuced into the film to correct for such, or add saturation

4) at this point you have pure dye transmission, add to these
transmittances, the transmittance of the aim optical printer used in the
design of the film, at aim filtration or exposure setup depending if it is
subtractive or additive

5) process these transmittances through the channel independent
sensitometry and interimage matrix of the print media

6) at this point you have pure dye transmittances for the print media's
dye, and using a transmittance to CIE-XYZ model for the dye, you can
encode the colorimetry of the print media

7) apply a colorimtery to appearance model for the XYZ you have since you
will need to consider the viewing conditions of the print media and want
to encode appearance rather than colorimetry in an ICC profile


to appearance onto a color negative film with a film recorder, ask for the
print media's appearance, by building an ICC profile this way


1) know the spectral exposure characteristics of the recorder
2) cascade it through the spectral sensitivity of the film
3) take the cascaded spectral transmittances through the channel
independent sensitometry and interimage for the film
4) process the transmittances through the spectral characteristics of the
film's dyes
5) now through the aim optical printer (at aim setup) transmittances
6) onto the spectral sensitivity of the print media
7) take the cascaded spectral transmittances through the channel
independent sensitometry and interimage for the print media
8) process the transmittances through the spectral characteristics of the
print media's dyes
9) apply a colorimtery to appearance model for the XYZ you have since you
will need to consider the viewing conditions of the print media and want
to encode appearance rather than colorimetry in an ICC profile

 

[MoiMoi]

to make a profile for a color negative scanner:

[steps 1-7]

to appearance onto a color negative film with a film recorder, ask for the
print media's appearance, by building an ICC profile this way

[steps 1-9]

See?
It's simple!

MM

 

[Noons]

color negative films are designed to make good prints, either on
photographic paper, or motion picture print film

- there is no "scan only" color negative film designed for accurate scene
reproduction
- there is no "film recorder" color negative film designed to work
specifically for a film recorder

while the above are ideals considered in the design of the film, the main
consideration is a good print, and thus the ICC specification for scanning
a color negative film is the characteristic intent of the film when it is
printed
snip

I've read what you said and it is very interesting: thanks for sharing
that.
On the first point above though, I have to slightly disagree: the new
Fuji Pro color neg films are as good for scanning as anything I've ever
seen.

To my taste at least, they are the best color neg films yet for
scanning.
That is with a good quality lcd/ccd scanner, others might see different
results?

 

[Tom Lianza]

color negative films are designed to make good prints, either on
photographic paper, or motion picture print film

- there is no "scan only" color negative film designed for accurate scene
reproduction
- there is no "film recorder" color negative film designed to work
specifically for a film recorder

Actually Dale, the modern Portra and Supra films from Kodak were
designed with Scanning in mind. I will agree that most of the real
efforts in this area were aimed at reducing artifacts from the
anti-halation layer, but nearly all negatives are scanned before being
printed, today. If you examine the Kodak literature on their sensors,
you will note that they have also shifted the sensitivity of the sensor
to accomodate the issues for film negative scanning.

while the above are ideals considered in the design of the film, the main
consideration is a good print, and thus the ICC specification for scanning
a color negative film is the characteristic intent of the film when it is
printed

As a member of the ICC, I have to say that I have never such mention of
negative film in any specification. The specification states that the
PCS is print relative with a limited dynamic range of approximately
280:1, with a known viewing condition, and know there is a reference
gamut that one can use for gamut mapping. That certainly DOES NOT MEAN"the characteristic intent of the film when it is

printed"

The most fundamental problem with Negative material is that capture
range of the film typicially exceeds the average scene range and the
output range (in density) exceeds the input log-exposure range of the
media that it is printed on ( for reflection printing). A simple review
of the characteristic curve data from Kodak or Fuji will show that.
Under these circumstances, the profile becomes very scene dependent as
well as film dependent. This is a similar problem to the issues we are
finding in digital camera profiles. The fundamental question is how do
you define the diffuse white (i.e. paper white in the PCS) and how do
you scale that with respect to the specular. This is especially
important if you are going to try to use appearance mapping. The 7 step
algorithm you described might get you to some reasonable estimate of
Scene Colorimetry, but as you probably know, that will produce an
unusable image for most aesthetic applications.

If you are going to use an appearance algorithm such as CIECAM2000, you
have to know the absolute scene luminance, which means that you need the
ISO-EV information from the original exposure. If you know the film,
you know the ISO value, but unless you have some source of scene
meta-data, you don't have the EV information and hence, you can't get
back to absolute scene luminance.

Even if manufacturers made the data you requested public, you would not
have enough information to make a usable robust profile.

regards,
Tom Lianza
X-rite Inc.

 

[Dale Kelly]

Actually Dale, the modern Portra and Supra films from Kodak were
designed with Scanning in mind.

you can only hope to minimize artifacts of spectral sensitivity to make an
accurate capture and need no further processing in scanning, that is
really all that you can do, and considering all sensitivity dyes are broad
spectrum, and have unwanted absorptions thereof, you cannot make a
scan-only film, the best you can do is use a transparency film

 

[Dale Kelly]

If you are going to use an appearance algorithm such as CIECAM2000, you
have to know the absolute scene luminance, which means that you need the
ISO-EV information from the original exposure. If you know the film,
you know the ISO value, but unless you have some source of scene
meta-data, you don't have the EV information and hence, you can't get
back to absolute scene luminance.

with a digital camera or scanner you can calculate scene luminance

for a digital camera, unbuild the gamma and filtration back to linear CCD
values then apply the L* algorithm

for a scanner,, unbuild the system

1) undo the film's dye characteristics
2)) undo the interimage and sensitometry of the process
3) undo the unwanted absoorptions of the spectral sensitivity of the film
4) at this point you have intensity and can apply the L* algorithm

 

[Dale Kelly]

As a member of the ICC, I have to say that I have never such mention of
negative film in any specification

last time I read it, it was in there

 

[Tom Lianza]

with a digital camera or scanner you can calculate scene luminance

for a digital camera, unbuild the gamma and filtration back to linear CCD
values then apply the L* algorithm

for a scanner,, unbuild the system

1) undo the film's dye characteristics
2)) undo the interimage and sensitometry of the process
3) undo the unwanted absoorptions of the spectral sensitivity of the film
4) at this point you have intensity and can apply the L* algorithm

Dale,

With all due respect, without the exposure time, you can't get back to
scene luminance, and you can't really apply appearance mapping. L* is
not an "appearance" model in the contemporary sense of appearance
mapping. You might take a look at Mark Fairchild's book on the subject
for a better understanding of the issue.

Regards,
Tom

 

[Tom Lianza]

you can only hope to minimize artifacts of spectral sensitivity to make an
accurate capture and need no further processing in scanning, that is
really all that you can do, and considering all sensitivity dyes are broad
spectrum, and have unwanted absorptions thereof, you cannot make a
scan-only film, the best you can do is use a transparency film

Actually, a transparency film has far greater issues with unwanted
absorptions than negative media. And I certainly hope that you are not
suggesting that getting back to scene colorimetry is somehow easier with
transparency film than it is with negative film.

Regards,
Tom.

 

[Dale Kelly]

Actually, a transparency film has far greater issues with unwanted
absorptions than negative media. And I certainly hope that you are not
suggesting that getting back to scene colorimetry is somehow easier with
transparency film than it is with negative film.

you don't have to get back to scene colorimetry anyway, I was just showing
that you could, the algorithm for appearance could include the overall
luminance and surround considerations of the substrate you are trying to
reproduce, not the scene, doing this with a chrome is much easier since in
using chromes the chrome is the object to reproduce, in which color okays
are made, you have no direct thing to reproduce with a negative, you
either build it to its print definition, or unbuild it to its scene
definiton

 

[Dale Kelly]

With all due respect, without the exposure time, you can't get back to
scene luminance

you can easily get back to get luminance by unbuilding the capture system
you use, a capture system is just as good of a meter than anything,
spectral measurement devices entail the same sorts of errors

 

[Tom Lianza]

you can easily get back to get luminance by unbuilding the capture system
you use, a capture system is just as good of a meter than anything,
spectral measurement devices entail the same sorts of errors

Dale,

Exposure is the product of illuminance * exposure time (lux seconds). I
can take a photograph for 30 seconds in the light of a full moon and
reproduce the same density on the film that I would in the daylight at
noon at 1/1000th of a second. That is the basis of the reciprocity law.
The two scenes ,when printed and viewed will look very similar,
however our visual assessment of the original scenes will be completely
different. That is what appearance models are all about.

A good starting point for understanding exposure can be found at

http://en.wikipedia.org/wiki/Exposure_value

If you read the section titled Relationship of EV to lighting
conditions, you will see the basis for the basic light meter equations.
You cannot unbuild luminance or illuminance from film density
without knowing the exposure time.

The metadata contained in most modern digital camera data provides
enough information to get to average scene luminance. It contains the
ISO rating, the f# and shutter speed of the shot. That didn't happen by
mistake.

I work in both the motion picture and still imaging fields. I'm working
with the ICC on developing a "minimal" data camera profile requirement
that is useful for ICC workflows. Trust me, this is not a trivial issue.

Tom Lianza
Director, Motion Picture and Video Technology Development
Xrite corporation .