Digital BW, The Print

> I agree with Austin.   Tell us precisely what information is
 > missing.  Don't speak vaguely about the "curves" because we know what
 > the response curve is of the color dyes used in the film.   So we
 > know exactly how much to compensate the density by for any color.
 > Or put another way:  because we know the shape of the film's response
 > curves there is only one unique point on the spectrum that will
 > produce a given density in all three dyes.  So what's missing?
 >
 > You say that other people here understand you, but I don't see them
 > jumping in to clarify what you're saying.

I think I need to weigh in again.  Perhaps no one read my previous post. 
*I* understand Anthony, at least on this point:

It is a basic mathematical fact that the process of binning is *not* 
reversible under almost all circumstances.

Let's say that we have a spectral input of amplitude A(F), where F is the 
frequency.  Let's take two simple response functions (the same applies to 
three, but I the extra lines clutter things up), r(F) and g(F).

If we take two binning functions to collect r and g...
R = integral over all F of (A(F)*r(F))
G = integral over all F of (A(F)*g(F))

Look at those last two equations.  If I give you the result of the 
integral (the R value, for instance) and the r(F) response function of the 
filter/CCD combination, can you tell me what the input A(F) is?

Answer: for all but the most trivial A(F) and r(F), no.  For an arbitrary 
A(F), never.

More simply:  if you know the exact spectral response of, say, the red 
element of a CCD sensor, and I tell you I got a red value of 130, can you 
tell me the original spectral input?

Answer: no.

-- 
Jon Dubovsky ( entropy@... )