Re: ICC profiles and digi-cam-last post

Lee Varis

Hi again,

To be fair to Andrew I think I need to clarify a few things,

I wrote:

White balance is kind of like applying a profile to the "raw" data.
to which Andrew responded:

I don1t agree because I see a huge difference between an input profile and
the role of white balancing. Again, profiles describe color, nothing more.
White balancing affects the response of the color.
OK, I said "kind of like" not exactly like.. but Andrew is right a
profile doesn't change the data at all only what the data means. White
balance affects how the raw data of the capture gets converted into some
meaningful RGB - the problem is that many people use custom profiles for
16 bit raw files to perform the same white balance ( now it is true that
it is only the appearance of the color is balanced, the actual RGB
values for neutral might end up being something like 124, 135, 129)
This is kind of dangerous IF you end up using the input (camera) profile
as a work space because it becomes much harder to evaluate things based
on the numbers in the info pallet. You must then convert from this input
profile to the R=G=B workspace - "kind of like" white/gray balancing.

I wrote:
all digital cameras will perform as well as film does in delivering
neutral color for most scenes even without icc profiles.
Andrew responds:

Because again, the role of the profile isn1t to set or alter gray balance.
It1s only to describe what the camera saw at any given moment. That1s why
you do want to gray balance and then build a profile. That1s why after
building the profile, you want to gray balance for each scene. This
accomplishes two things. It insures gray balance (which sometimes you don1t
want but often do) and it places the capture into a condition that matches
how the original capture was produced to build that profile. So with gray
balance, the profile is valid. The profile will have NO role of the
neutrality of the capture. The profile only records a condition. You have
have a file with a gray and Assign or not assign a profile and the RGB
numbers of the gray will not change a lick.
Andrew describes fairly succinctly the best profiling practice for
digital camera captures here. You certainly will not hurt your chances
for getting good color if you follow this approach. To be fair there are
some cameras which seem to require this approach due to some
idiosyncratic color conversions from their raw data (the Nikon D-1 and
the Kodak DCS 460 come to mind) Again, MOST cameras will deliver a gray
balanced 8 bit RGB where R=G=B when "processed" from the native "raw"
file into an 8 bit Tiff

I wrote, concerning the gray balanced RGB from most camera software:

Once this is done a camera file can be
opened up in Photoshop and assigned any of the standard working spaces
and give you a very reasonable starting place for color.
to which Andrew responds:

No! It will only insure you have a neutral gray. If what you say is true,
you could capture the scene as you suggest and Assign sRGB and Wide Gamut
RGB and you1d get to the same place.
Actually, what I am suggesting here is that you can easily evaluate
general color of the file visually simply by assigning a workspace
profile. Now if you choose sRGB the image may look a bit desaturated or
it may look OK. I doubt that "Wide Gamut RGB would look anything but
over saturated but I hardly consider Wide Gamut RGB to be "standard"
even though technically it is one of the "standard colorspaces in
Photoshop. If you pick sRGB, ColorMatch RGB, Apple RGB or Adobe 1998 the
displayed color won't be radically BAD because all of these spaces are
similar to monitor RGB. Andrew is right however in saying that they
won't be in the same space. They will be "in" whatever space you assign.
I apologize if that wasn't clear. I suggest you try a number of spaces
and pick the one that looks the best to you. If you then adjust the
color to your liking you won't be surprised when you convert to the
desired output space.

Andrew elaborates:

Nothing could be further from the
truth. Assigning a profile only describes to Photoshop 6 the MEANING of the
numbers in the file. As you assign different profiles, the numbers NEVER
change. But the preview (and any further conversions FROM the assigned
profile) will change. IF you have a file that is reasonably close to say
ColorMatch RGB, assigning it ColorMatch RGB will work pretty well. Assigning
any space that moves further from ColorMatch RGB will hose the file (Preview
and conversions) farther and farther from the true meaning of the data.
I tend to be less dramatic in my assessment of the necessity for
"correct" interpretation of the "true meaning of the data" but basically
Andrew is right! I've also found that most digital camera mfgrs. assume
that you are evaluating colors on a monitor and they render their
digital camera captures to look good on a calibrated monitor - that
means the color will be reasonably close to ColorMatch RGB! I've had
good luck with Adobe 1998 as well as sRGB - it kind of depends on the
image. 3D gamut charts aside, these three colorspaces are not that
radically different from each other.

Assuming a digital camera file is in ColorMatch may work or may fail. But if
you actually profile the camera, you know the exact meaning of the numbers.
Realize too that the Working Spaces you can Assign are not based on any real
device. They are synthetic RGB models that work well because they all have
R=G=B as a neutral. You can actually gray balance a digital camera file and
find that this raw capture has data where R, G and B do not equal UNTIL you
convert into the Working Space! Input colorspaces do not insure R=G=B yet
you still have a neutral.
Here is where my experience seems to differ. When I gray balance the
digital camera file using the camera mfgrs. software the numbers (actual
RGB values) end up R=G=B and they stay that way when you open them in
Photoshop (if no conversion occurs) if you use the camera software to
modify the raw data (which is what you do when you save out an 8 bit RGB
file). Now if you really desire to use the raw 16 bit linear data from
the camera and open this up in Photoshop you will need a custom profile
to describe the color that the camera saw otherwise you will be pretty
lost. I mentioned this before:

Neither of these approaches
will be quite as good as using a VERY good "custom" profile for the 16
bit "raw" data file but both of these approaches are much more practical
for the vast majority of images you need to capture.

I don't generally find it necessary to work with raw 16 bit data for
most images.

I wrote:

The benefits are not great enough to
warrant the extra trouble either - it is expensive and time consuming to
get a custom profile that would be better than simply assigning a
workspace to a white/gray balanced file for even 50% of the subject
matter you are likely to photograph.
Andrew responds:

Depends on the camera and what the raw data is. Be my guest and assign any
Working Space to a Nikon D1 image and I assure you they all look like crap.
...Raw D1 RGB isn1t anything like any of the
supplied RGB Working Spaces. You can1t assume that an input device is
creating RGB that is in any way close to an RGB Working Space you might
Andrew is again correct as far as "raw" data is concerned and certainly
the Nikon D-1 is fairly goofy if you simply interpret the "raw" data as
an RGB work space. I really wasn't talking about "raw" data though. All
of the cameras I have worked with will deliver an RGB "processed" file
in 8 bits which behaves fairly reasonably if you assume an RGB workspace
- any deviation from ideal can be handled with a Photoshop correction
that you can save and use for every file you open from the camera. There
is a theoretical advantage to delaying and minimizing color conversions
by using input profiles with camera files and that is what Andrew is
championing here and I can't find fault with that approach. I don't
bother with this myself for commercial work and it seems to work fine -
there's always that little nagging feeling that it could be a little
better though so if it helps you sleep better at night...

I write...

For instance, if you manually white balance a sunset scene you will "balance
out" the overall red color and completely destroy the feeling of the
sunset in the image. It is the same thing with profiles.
Andrew responds:

No, input profiles don1t change the data, they only describe the data. IF
you built a profile assuming a white balance, you have to white balance.
That act of white balance hoses your sunset, not the profile.
OK, he got me again.... Yes, profiles don't change the data they only
describe the color. White balance is not exactly like a profile because
it changes the data! Now if you are foolish enough to shoot a Gretag
Macbeth DC target in sunset light to build a profile for sunset lighting
you will hose the sunset because you will normalize the interpretation
of the color to the red sunset light - neutralizing the red even though
you don't change the data!

I write:

I think the challenge for photographers or anyone dealing with
digital photographic images is to create great images by interpreting
color in the most emotionally satisfying way.
Andrew responds:

How do you interpret 11s and zero1s? That1s all a digital file is. That is
why Photoshop 6 needs the proper profile to be assigned. Now you have
properly interpreted the data for Photoshop 6 to provide a correct preview.
You1ve got the meaning of the numbers necessary to convert to a Working or
Output space. Once you take your 3great2 image and Assign the wrong
profile, it1s not so great anymore, at least visually.
Well... I don't really interpret one's and zero's I look at the image on
my monitor and decide if I like it or not. If you assign an RGB
workspace to any RGB file it may not be technically correct or look very
close to the original physical object but it will look like SOMETHING
and it has been my experience that it's not that hard to make it look
like what you want it to look like. If you can't decide what you want it
to look like then a profile may provide some some feeling of security
but not having a custom input profile is not going to be such a disaster
either. Photoshop does not care if you assign the "proper" profile -
even if you do not assign any profile Photoshop will simply assume
whatever workspace you've set for the default and present you with an
image that you CAN evaluate and interpret.

Andrew is correct in saying that if you have a good profile for your
camera you can have a good preview of the color on your monitor and this
would be ideal because you could minimize the actual color transforms to
get your image into a printable form. What I am saying is that to get
great color you WILL have to apply some judgment and creative control
over the rendering of that color whether it utilizes a custom input
profile or not. You can get good color without using custom input
profiles as long as you are properly utilizing the other aspects of a
color managed workflow. You MUST have a reasonably good profile for your
monitor IF you are going to be evaluating colors visually. You MUST have
a good profile for your desktop inkjet printer IF you are going to be
doing any cross-rendering (like simulating a press on your Epson
printer) or softproofing to the monitor. These activities make producing
good color easier with digital camera images because there is no
physical point of reference for color like a piece of film. If you do
not have a calibrated/profiled monitor then you are going to have a
really hard time interpreting anything very meaningful from one's and
zero's (not impossible though)

I'm going to end here by saying that custom input profiles for digital
cameras are NOT A BAD THING. In fact, in some circumstances, they can be

Let me say, however, that you CAN get by without them WITHOUT monumental
effort though you may derive a measurable advantage in using custom
input profiles if you are having a lot of trouble with your camera's
color. I have not found a camera yet that produced files that I couldn't
work with without resorting to custom profiles.

Let me also say that Andrew knows a lot more about profiling digital
cameras than I do and I believe that he can profile a camera to render
better color out of the box than I can get without using a profile - I
will have to make some corrections to be better than his first try and
this could be very valuable to someone who shoots catalogs of hundreds
of shots. I also believe that in the next couple of years this will get
much easier to do. I also am looking at a number of different profiling
options myself because I am convinced that there are some creative
applications for the use of ICC profiles. I am hoping that I don't have
to purchase a $3000 package to get profiles that I will have to edit
anyway - Photoshop does a lot more than any profiling package and it
only costs $900.

OK Andrew, I'm bracing for your one-two knockout punch......


Lee Varis

Join to automatically receive all group messages.