Re: [CinePaint] Working with RAW data

On Sun, 23 Mar 2008 23:19:53 -0500 (CDT)
Bob Friesenhahn <[email protected]> wrote:

> 
> The phrase "the whole issue can be quite complex" implies that there 
> is some sort of science detailed in this conversation thread which 
> goes beyond "works for me".  It seems best to simply admit that using 
> ICC profiles in this way is a "works for me" type situation.
> 

My feeling is that the issue may be a bit too involved to rehash
on the Cinepaint mailing list.  Thus I made an attempt to refer
the original poster to another source of information.

Proprietary software does an excellent job of processing
raw files without the need for custom color profiles.
How is this done?  The UFRAW mailing list is, as far as I
know, one of the only places where this issue is considered,
and those discussions would certainly be relevant to Cinepaint
users.


> While CCD and CMOS sensors are fairly simple and easy to characterize, 
> as used in consumer digital cameras, there is firmware added which may 
> make changes to the biasing of the sensor before the shot is taken, as 
> well as adjust the iris and shutter duration.  The sensed colors have 
> quite a lot to do with the colors surrounding the object and the 
> overall color temperature of the scene.  The zoom level of the lense 
> has some impact on the appearance of colors.  The camera is likely to 
> exhibit a degree of metamerism.  The Bayer pattern does not help since 
> it invents colors and sensor sites which do not actually exist. 
> 

Those are all valid points, but the fact is that although these details
and parameters should be published by the camera manufacturers they
are not.  Consequently the user is left guessing.  Custom color
profiles are certainly an answer to all of the uncertainty, but as I will
explain below, they can be more trouble than they are worth.

>
> There are a large number of variables which are unlikely to be 
> normalized by a gross "white balance" adjustment done by software like 
> 'dcraw'.
>

Actually, it is more than just a gross adjustment.  White balance
data specific to a particular scene is stored by the camera firmware
in the raw file.  Dcraw uses that data to adjust the color in the same
manner as does proprietary software.  The combination of this
white balance data plus a generic color profile supplied by the manufacturer
is usually sufficient to faithfully reproduce colors across a wide
range of external conditions.

The alternative is, as you indicate, to prepare a color profile for
each unique situation.  In the absence of a manufacturer supplied
profile there would be no other choice.  But for the active
digital photographer, this would necessitate creating a profile
for every hour throughout the day, and for every day throughout the
seasons, to match all possible lighting effects, including the presence
and absence of clouds, shade, etc. Nothing short of hundreds, if not
thousands, of separate and distinct color profiles would be needed.
To eliminate this madness, the white balance multipliers are utilized
to normalize the data.

For professional use, under controlled studio conditions employing
highly reproducible illuminants, the creation of individual
profiles may be desirable and feasible.  But for anything else
it would be a ridiculous undertaking.

Proprietary software does process raw files in the manner described
above using the white balance and a single standard input profile,
or at least a limited range of profiles.  The details will depend on
the particular manufacturer, and none are very forthcoming about
these details.

As an example, I have had great success in processing Nikon raw
files using Nikon-supplied generic profiles.  A single profile
will work regardless of the type of illumination.  Indoor tungsten
lighting will process just as faithfully as will outdoor scenes
using this profile due to the prior white balance adjustment.

With Canon raw files, however, I have had poorer results.  This
could be due to the lack of suitable profiles -- I have not yet
discovered the exact profile that is used by Canon -- or it could
be something else.  Raw processing involves many steps, and usually
there is some type of gamma correction performed before the
application of the profile, but the nature of this transform
is not published.  Consequently, one has to experiment in order
to get acceptable results.

It is important for the user to understand both sides of the
story.  Individual color profiles can be tailored to suit a
specific situation or perhaps all situations, but the labor
involved would be extreme.  The far easier route for most
cases would rely on the white balance normalization and
the use of a standard profile, if obtainable.

FP


-------------------------------------------------------------------------
This SF.net email is sponsored by: Microsoft
Defy all challenges. Microsoft(R) Visual Studio 2008.
http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/