filmscanners: Vuescan color spaces (long)

["shAf" <michael@shaffer.net>]

     ... having asked you to "stay tuned", I'm getting back to you.
However, what I've experimented with has led me into considerable
thought ... my apologies for this post's length  :o)

    First ... my conclusions remain the same ... Vuescan will
ultimately yield more color capacity (gamut) than most of you will
ever need.  My experiments imply a gamut for which Ektaspace, AdobeRGB
and all others (except ProPhotoRGB and Adobe Wide gamut) are very
viable working spaces subsequent to scanning with Vuescan.  Vuescan is
unique thanks to Ed Hamrick's philosophy for scan control and
providing for such experimentation.

    What is more ... Ed informs me there is no such internal color
space as "VS RGB"... Vuescan simply carries scanned RGB from the
device to any of the selectable "color spaces".  The scanners are
characterized, recognized if on line, or selectble from the
"device|mode" list if scanning from a raw file.

    Yet, there is room for improvement.  In the same sense we would
all like to sell our 2700dpi scanner and purchase 4000dpi, we are
ignoring another property of increasing the detail ... that being
COLOR.  On one hand, I claim VS provides you with plenty of color ...
but I do see evidence for improvement.  Bruce Fraser, in different
contexts, has stated Ektaspace is generally capable, but he also
qualifies with actually seen real life examples where the gamut
exceeds Ektaspace.  Therefore, we might ask more of scanning software
for color, ... analogous to asking more of the hardware for DPI
resolution.

    Skip this paragraph if you know what I'm talking about.

    There are two types of color profiles ... "device dependent"
(generally referred to as "device profiles"), and "device independent"
(generally referred to as "working color spaces" ... e.g., AdobeRGB,
sRGB).  Working color spaces are simply defined ... R-G-B primaries,
gamma and white point ... converting from one to another can be
reduced to a simple mathematical matrix operation.  Device profiles,
on the other hand, are complex ... said to be 3-dimensional ... or
look-up table defined ... no mathematical transformation is simple.

    Photoshop v.6 has a new capability ... it will simplify a 3-D
device profile to a supposedly equivelent color space.  For example, I
have a 3-D LUT Nikon_wide device profile (identified by its large file
size, 480kb) provided with the Nikon software.  (We all take Nikon CMS
with a grain of salt, but maybe the person who created this profile
knew what he was doing and the NikonScan software people did not.)  In
any case ... when I assign this profile to a Vuescan "color
space=device RGB" image, it becomes obviously oversaturated.  ( ...
ok, ok ... assume I've done something totally inappropriate ... read
on ...)

    Recall, PS6 will "simplify" a device profile.  If I do this with
"Nikon_wide", PS6 saves "simplified Nikon_wide" ... and it is indeed
simplified ... it is now 1kb in size.  Now, what happens if I assign
this simplified profile to my "color space=device RGB" image???  It is
NOT over-saturated and it looks identical to my Vuescan "color
space=Ektaspace" image!!!

    What does this tell us?  For one, it implies a simplified device
profile is nowhere equivelent to the original.  It also tells "me" (if
I believe my Nikon supplied device profile), I am not capturing all of
the gamut capable from my scanner with Vuescan.  It also implies
Vuescan impliments scanner profiles at the "simplified" level.  BUT
... before you ask Ed for improved color gamut in version 8, be
prepared to ask Ed for a lot of work AND for most likely paying him a
lot more than $40.  Ed will likely have to license Adobe Color Engine
or the Kodak equivelent ... not to mention hire PR and support
personnel.  It would mean in my mind "VS Pro".

    Lastly ... don't get me wrong ... even if I don't get what I want,
I'll continue using VS ... that is, I won't get it elsewhere ... even
with Nikonscan or Silverfast.  I also don't know what I'm talking
about with respect what "VS Pro" may demand from Ed ... all I know is
how to experiment with the fundamentals of color spaces, "working" and
"device dependent".

shAf  :o)