Digital BW, The Print

Peter writes:

> Let's not get all mystical about it.

It is not mysticism, it is hard science, as I have explained ... several
times.

> A response curve is a vector
> of scalars.

Well, that vector has only three scalars in it after you take the picture:
red, green, and blue.  Try to reconstruct an arbitrary curve from that.

> It should be simple linear algebra to take two
> response curves and come up with a third curve
> that represents the conversion of one to the other.

Sure.  The problem is that you have nothing to which you can apply that
third curve.  You have only three values: red, blue, and green.  You need a
whole curve to which you can apply your conversion--but it's gone.

> The only case that would fail is if the
> target response curve had a wider latitude than
> the source one.

_Most_ cases will fail.  Unless the entire curve is a function of the three
RGB values, it will fail.

--- In DigitalBlackandWhiteThePrint@yahoogroups.com, "Anthony 
Atkielski" <anthony@a...> wrote:
> Peter writes:
> 
> > Let's not get all mystical about it.
> 
> It is not mysticism, it is hard science, as I have explained ... 
several
> times.
> 
> > A response curve is a vector
> > of scalars.
> 
> Well, that vector has only three scalars in it after you take the 
picture:
> red, green, and blue. 

No it doesn't - it has however many scalars you need for your bit 
resolution.    For a 12 bit scan it has 4096 scalars in it, and 
remember, you have three vectors - one each for the R,G, and B 
response curves.

Peter writes:

> No it doesn't - it has however many scalars you
> need for your bit resolution.    For a 12 bit scan
> it has 4096 scalars in it, and remember, you have
> three vectors - one each for the R,G, and B
> response curves.

Feel free to answer my question about the yellow-sensitive film, if you
believe I am wrong.