Digital P&S and color bit depth

Raphael Bustin writes:

And just to be clear, if it's a Nikon film scanner like
the LS-8000/9000, it's a *monochrome* CCD with
no filters at all over it. Instead, the Nikons use three
sets of LEDs -- red, green and blue, and alternate
these rapidly -- just as in a Canon LIDE flatbed scanner,
though the Canon uses a CIS sensor, rather than CCD.

Good point; I neglected to mention this way of handling the colour.

It can work really well when scanning film, because you know that the
film image is really made up of three dyes, so the ideal light source
for digitizing it is three pretty narrow-band (pure colour) sources.
The manufacturer would choose the wavelength of the red light source so
it is strongly absorbed by the cyan dye, but almost unaffected by the
yellow and magenta dye, and so on for the other two colours. This
minimizes crosstalk between colours. (It also makes it easy to add a
4th measurement channel using IR light for "ICE" processing).

On the other hand, it's not so good for a flatbed scanner, which is
supposed to reproduce colour accurately for anything you put in it.
There, you really want the light source and sensor filters, taken
together, to have a response more like the human eye colour-matching
functions, and that eliminates single-frequency light sources. (The
difference is that with a film scanner, there *are* only 3 colours in
the film that you care about, while with real-world objects you have the
whole visible spectrum to consider).

(Despite that, the LiDE scanners use a monochrome sensor and RGB LEDs.)

Film scanners *do NOT* use Bayer sensors.

Nope, there's no need to. Bayer sensors show up only where you *have*
to capture the whole image at the same time. A slide scanner can take
its time building up the image one line at a time. (This also has
advantages in reducing lens flare, since the light source can be a line
instead of the full frame area).

Dave

On Mon, 30 Oct 2006 18:17:05 +0000 (UTC), (Dave
Martindale) wrote:


It can work really well when scanning film, because you know that the
film image is really made up of three dyes, so the ideal light source
for digitizing it is three pretty narrow-band (pure colour) sources.
The manufacturer would choose the wavelength of the red light source so
it is strongly absorbed by the cyan dye, but almost unaffected by the
yellow and magenta dye, and so on for the other two colours. This
minimizes crosstalk between colours. (It also makes it easy to add a
4th measurement channel using IR light for "ICE" processing).

On the other hand, it's not so good for a flatbed scanner, which is
supposed to reproduce colour accurately for anything you put in it.
There, you really want the light source and sensor filters, taken
together, to have a response more like the human eye colour-matching
functions, and that eliminates single-frequency light sources. (The
difference is that with a film scanner, there *are* only 3 colours in
the film that you care about, while with real-world objects you have the
whole visible spectrum to consider).


FWIW, according to Mike Chaney (the designer of Profile
Prism, a print-profiling package) the Canon LIDE scanners
are work well with his scanner-based profiling package.

Presumably this would be another benefit of the narrow-
band illumination? And now that I think of it, does that not
(in general) contradict your claim? After all, most prints are
also made with CMYK dyes or pigments (though these
may not be as narrow-band as the dye layers in film)


rafe b
www.terrapinphoto.com

On Mon, 30 Oct 2006 18:17:05 +0000 (UTC), (Dave
Martindale) wrote:

It can work really well when scanning film, because you know that the
film image is really made up of three dyes, so the ideal light source
for digitizing it is three pretty narrow-band (pure colour) sources.
The manufacturer would choose the wavelength of the red light source so
it is strongly absorbed by the cyan dye, but almost unaffected by the
yellow and magenta dye, and so on for the other two colours. This
minimizes crosstalk between colours. (It also makes it easy to add a
4th measurement channel using IR light for "ICE" processing).

On the other hand, it's not so good for a flatbed scanner, which is
supposed to reproduce colour accurately for anything you put in it.
There, you really want the light source and sensor filters, taken
together, to have a response more like the human eye colour-matching
functions, and that eliminates single-frequency light sources. (The
difference is that with a film scanner, there *are* only 3 colours in
the film that you care about, while with real-world objects you have the
whole visible spectrum to consider).

Raphael Bustin writes:
FWIW, according to Mike Chaney (the designer of Profile
Prism, a print-profiling package) the Canon LIDE scanners
are work well with his scanner-based profiling package.

Presumably this would be another benefit of the narrow-
band illumination? And now that I think of it, does that not
(in general) contradict your claim? After all, most prints are
also made with CMYK dyes or pigments (though these
may not be as narrow-band as the dye layers in film)

Photographic prints have just 3 channels of information, represented by
3 dyes. 4-colour ink printing also has just 3 channels, really - the
black ink makes up for limitations in the 3 colour inks. So 3
narrowband sources might well give good results scanning these subjects,
particularly if you can profile the scanner.

On the other hand, suppose you place a tree leaf showing its fall
colours into your flatbed. I think that will have a wider range of
colours present, and there you're likely to see more colour errors if
narrowband light is used.

Dave