Digital Camera: Raw Data + ICC profile = one very bad idea....
Digital Camera: Raw Data + ICC profile = one very bad idea....
- Subject: Digital Camera: Raw Data + ICC profile = one very bad idea....
- From: "tlianza" <email@hidden>
- Date: Mon, 24 Feb 2003 07:50:54 -0500
Hi,
I've been reading some comments about Raw Data and ICC profiles and I
thought that I would make some comments here in that regard. As the subject
heading indicates, I think that using custom input profiles with digital
cameras is a bad idea EXCEPT WHEN IN A STUDIO SITUATION. I'm sure there
will be some disagreements on this point and I really encourage those who
have been successful with input profiles in the field to share their
experience with examples. I've been using a Nikon D100 with some of the Raw
capture programs, plug ins etc, and I have very mixed feelings about the
efficiency of the workflow. I've only take a few thousand images, so my
experience is somewhat limited. With regard to color, especially out in the
field, I really don't see the benefit of a profile. Let me make some points.
1. The camera doesn't have a gamut limitation.
A well designed camera has overlapping sensitivity functions and ,BY
DEFINITION, as long as those sensitivity functions match or exceed the
limits of human vision, within the dynamic range of the camera, the visible
color space is captured. The same thing is true with film, in general. It
may not reproduce the color correctly, but all naturally occurring colors
are there. The gamut "limitations" occur when a color space is selected. In
the Nikon there are two sRGB models and one Adobe RGB selection. The reason
for two different sRGB models is that each setting has it's own bias. One
is to be used for landscapes and one for portraits or flesh tones. Nikon is
emulating the same bias that a photographer might use in selecting a film
type. A lot of color scientists snicker at the concept, but the fact is
photographers have preferences that don't necessarily rank absolute color
reproduction high in the chain. Another point that is often lost is that
camera manufactures also apply adaptation assumptions into the processed
data. In almost all cases, you will see that the tone reproduction is not
exactly what the workspace would imply. For instance with a camera working
to the sRGB space, the effective gamma of the reproduction is normally
higher than 2.2 . Why? Because the images look a lot better when viewed in
a low ambient environment. There is a great deal of art and empiricism that
exists in the input world of photography. The whole thought process is
completely counter to profiling. The tools may be similar, but the whole
intent is quite different.
2. Profiling a film-scanner combination makes perfect sense, profiling a
camera does not.
The photographer selects a film for a particular rendition. Rarely, if
ever, does this have anything to do with proper color reproduction. As a
matter of fact, most photographers, when out of the studio, select
film/filter combinations that alter the original color reproduction
considerably. Very often,
they will select a film based on it's saturation characteristics (like Fuji
Velvia) and then shift the overall color balance with a warming filter. We
color scientists, profile the film-scanner combination to reproduce that
photographers vision on the film. That's exactly what our role should be:
Preserve the Artists Vision. Profiling a camera in the field, takes the
"vision" out of the process. As a photographer, I can tell you that's not
what I want to do. More importantly, I firmly believe that most of my color
science brethren don't have a clue about the background studies that go into
development of particular film characteristics. All of that work is
absolutely counter to the role a profile plays in the process. The factors
that make an image attractive are rarely reflected in a technically perfect
reproduction.
3.0 Profiling a camera outside of the studio is a fundamentally flawed
process.
In the late 1500's, DaVinci pointed out that all colors in a scene are
affected by the color of the light illuminating that scene. Leonardo
understood that "white" is just another color. At the heart of the ICC
concept is the Profile Connection space. The new ICC 4.0 specification has
really moved towards making this process mechanically feasible, but the
science is by no means complete. The problem is that the adaptation
required for the PCS involves not only a White Point adaptation,
but a luminance adaptation as well. Currently, there are four or five white
adaptation models that are useful under different circumstances and there
are at least that number of non-linear tonal response functions based upon
absolute luminance of the original scene. Adding the potential for over 20
combinations of "opinions" into the process hardly gives one the confidence
that any fundamental increase in quality can be achieved at this point in
time. The studio is a completely different case. Here, with strobe
lighting, the basic viewing conditions and the PCS align quite well.
Profiling in the studio makes absolute sense, if the goal is to reproduce
the colors accurately. A photographer should be prepared for some pretty
boring images however. A working photographer can simulate the results of
this process using film. Kodak makes a film called Ektachrome Professional
(EPN). It is very expensive, it renders color with extreme accuracy and
it's contrast is relatively low. It's the closest analog I can think of to
a color managed camera. If you do comparisons with this film and your
favorite, you will immediately see what I mean. The EPN consistently
reproduces the scene with far greater accuracy than the other films, but the
images look absolutely lifeless when compared to Velvia or the higher
saturation Ektachromes. As the relatively low sales of EPN would indicate,
most photographers are more concerned with the "Vision-thing" rather than
the "Color-thing" .
4.0 Color failures on a modern Digital SLR are generally well behaved.
The same cannot be said for a table based profile. The potential for
problems is huge, particularly in the blue area of the spectrum, where the
most of the color models simply don't work well. I would like to see some
color failures in field images presented by working photographers,who are
not using custom profiles. I'd really be interested in seeing results from
profiled cameras side by side with native reproductions. I'd be
particularly interested in a demonstration that shows that a profiled camera
works better than a non profiled camera in actual field use. While I'm
certain that improvements in the tone reproduction are possible, I have real
doubts that profiling can help the photographer get "better" color.
5. What would I recommend?
As a color scientist and a photographer, I would recommend capturing raw
data. If I had to build a general profile, it would map the data (using a
matrix-shaper profile) to a color space that was SMALLER than sRGB. The
gamma response would be based upon fitting the maximum dynamic range of the
camera into the new definition of the PCS (on the order of 288:1) to a
simple gamma law model. The image I would see on the screen would appear to
be low contrast and low
saturation, but all the underlying data would be there. If the rest of the
system was color managed, and I was working with the full range raw data, I
could make a cuves adjustment and a saturation adjustment in seconds, and
as long as the Monitor and Output profiles were accurate, my output results
would be
controllable and predictable. In my opinion, that's what the working
photographer is looking for: predictable and controllable.
I'd like to hear some discussion, pro or con.
Tom Lianza
Technical Director
Sequel Imaging Inc.
25 Nashua Rd.
Londonderry, NH 03053
email@hidden
_______________________________________________
colorsync-users mailing list | email@hidden
Help/Unsubscribe/Archives:
http://www.lists.apple.com/mailman/listinfo/colorsync-users
Do not post admin requests to the list. They will be ignored.