Re: How do I get a negative orange mask *back*?

Ewald R. de Wit (ewald@pobox.com)
Fri, 26 Nov 1999 21:31:55 +0100

Manuel Panea (mpd@rzg.mpg.de) wrote:
> Hi. Here is a question to the experts.
>
> I'm developing the Canon backend. The backend works with some Canon
> flatbed scanner models and with my 35mm-film scanner model 2700F. This
> model can scan negatives and slides.
>
> When scanning negatives, there is a SCSI command to set the scanning
> speed. In this way, dark photos can be scanned at a lower speed, exposing
> the film to more light so that the scanned image looks good. Alas, this
> speed setting command works only for negatives, not for slides (that's
> what the specs say and it's true, I've tried it). Unfortunately, the only
> available and fixed scanning speed for slides is invariably too fast.
> Almost all slides get scanned too dark. What to do?

You can multiply the data by a constant and then gamma correct it
(with gamma ~2.2). This will of course amplify the already abundant
noise of the scanner too. What it boils down to is that this Canon
is just not very good for dark slides.

Btw, I remember that the earlier 2700f models had a problem with being
too dark, if this is the problem perhaps you can return it. If you put
in an empty slide holder you should get 100% pure white, if not it is
just plain faulty.

> I thought I could try to outsmart the scanner and tell it to scan a
> negative at a low speed when in fact I am feeding it a slide. Inverting
> the colors is then an easy problem. The difficult problem is that, when
> scanning negatives, the scanner automatically removes the orange color of
> negative films. Slide films do not have this orange color, so if I try
> this trick, the scanner gives me an image where all information about
> orange-like colors is missing.

You could get better definition of the blue channel this way but then
the R and G channel will suffer. The orange mask of negatives absorbes
a lot of blue and to get a good blue definition your scanner uses a
longer exposure time. But the green and moreso the red is absorbed
much less and to avoid overflow of the R and G channel your scanner
will diminish the R and G signal (for example by dividing). This way
the full R, G and B bit ranges are utilized without getting overflows.

Now, if you put in a slide, you will get (for example) the full
range of the B channel, halve the range of the G channel and 1/3 the
range of the R channel.

> Does anyone know of a good way to get this orange back? Just "adding" it
> with e.g. the Gimp doesn't help, the image still doesn't look good. With
> some other operations, specially "equalize" I can get the image to look
> almost good, but not quite, and worst of all, the histogram is then "full
> of holes".

It is not difficult to uncover the mapping that the scanner does. The
HP Photosmart scanner for example, when scanning negatives, uses a
1:2:3 ratio for the R:G:B exposure.

-- 
  --  Ewald

--
Source code, list archive, and docs: http://www.mostang.com/sane/
To unsubscribe: echo unsubscribe sane-devel | mail majordomo@mostang.com