Digital BW, The Print

> I don't understand the relevance of that.  You do NOT have only the red
 > element (unless it's only red, that is), you have three values.  Also,
 > I do not see how knowing the exact spectral response of the CCD is
 > required to do this.  It IS one method, but, IMO, a poor one at best,
 > as it adds another variable that is unnecessary.
 >
 > What you do is simply create a "map" for the RGB values to grayscale
 > values using a LUT.  The LUT is generated by taking a full spectrum
 > image with both the B&W film, exposed and developed as you like and
 > the color film, scanning the two films on the same scanner, and
 > "matching" the graytones to the RGB tones at the same point in the
 > image.  Of course there is going to be some error, but I believe
 > the error is insignificant.
 >
 > I do not believe this needs to be calculated, it can be, if the
 > mapping is deterministic, what is called, characterized.  If scanner
 > differences are an issue, you simply characterize this for the
 > different scanner, following the same procedure.  If you do it once,
 > you can do it a thousand times, for any combination of films.

Except in the most trivial of cases, given the output of any number of 
binning functions, it is a mathematical impossibility to convert those 
into the outputs of a different set of binning functions.

Why?  The aggregation of A(F) into any number of outputs fewer than all of 
F results in a masking of A(F).  After that, it is impossible to 
reconstruct A(F).

Mathematically:  Beyond trivial response functions (i.e. an impulse 
function), multiple A(F)'s can produce the same outputs.  I.e. an output 
does *not* map to a unique input.

Finis.

-- 
Jon Dubovsky ( entropy@... )