Digital BW, The Print

Austin writes:

> What, precisely, does B&W film "sense"?

Light falling upon it.  Light in a wide range of frequencies will cause the
film to react, but the magnitude of the reaction per unit exposure varies by
frequency in a largely random way.  B&W film essentially integrates the area
under an irregular curve, and produces a single number as a result.

> Both the film, and the CCD, sense the number of
> photons that "hit" it.  They are spectral independent.

No, they are not.  Photons of certain energies produce a reaction more
readily than others.

> Look at the film response curve for B&W film.  It's reasonably
> flat, up to the point of fall-off.

It looks irregularly curved to me.  It can be dramatically changed by
filters, too.

> Now, does the RGB data not have this "intensity"
> information?

RGB data, like B&W data, is an integration of the area under one or more
curves.  The curves themselves are lost in the process.

> Does it matter that different combinations of RGB
> produce the same intensity?

What matters is the spectral distribution in the original scene.

I'll try again, although I'm not optimistic that anyone who hasn't already
understood will understand from this:

Set up two omnidirectional microphones in the middle of an orchestra, and
record some music.  At the same time, set up two other microphones, in two
completely different positions, and record the music with those, too.

Now take the recording from one set of microphones, and transform it into a
recording of the music as heard from the other two microphones.  You may be
surprised to discover that it cannot be done.  For example, an instrument
exactly equidistant from the first two microphones may be in one of two
positions, but that instrument will not be equidistant from the other two
microphones.  Therefore, the instrument will sound just as loud in both
channels of the first pair of mics, whereas it will sound unequally loud
across the two channels of the second pair.  Explain how you can transform a
recording from the first pair so as to accurately reproduce the imbalance in
the channels that was heard by the second pair.  It can't be done.

In fact, even if you use two mics for the first recording, and only one mic
for the second (like RGB to B&W), it still can't be done.