Hello!
These ae some of my observations (and wishes) concernig the
transparency adapter for the UMAX Astra 1220S.
When I scan a normally exposed slide, it comes out quite dark, and
I must stretch the histogram in XSANE (GAMA=1.3 B=120 C=120 for example)
This causes 'holes' in the output histogram (the dark values go 0, 7,
11, 15...). So it may be desirable to scan internally with 12
bits/color, even if the output is only 8 bits - to provide 'meat' for
histogram stretching.
Also, the dark areas tend to be quite noisy (same with twain/win).
It seems this is the CCD / readout electronics / AD converter noise.
Increasing the exposure time would help here (in the dark areas) but
would probably cause overflow in the bright pixels.
What I do now is to scan the same slide several times and average
the scans later in the application program. The results are well
worth the extra work!
So, I would like this averaging done in the SANE backend, like the
NIKON super coolscan 2000 (LS 2000) film scanner does it.
Again, this could be done with more bits internally. Actually, such
averaging can give you more bits of (real!) resolution than your A/D
has.
Tha averaging could be done two ways:
1. at each scanhead position take several readings and average them
before moving the scanhead on.
2. run several full scans (running the scanhead back and forth across
the scan area). (smpling only when moving forward)
This second way may seem stupid, since it needs more memory and the
mechanism inacuracies will reduce the spatial resolution, but it has
one nice property: because each scan is done at a slightly different
position, the film grain will average out.
I have found this to have a much better grain supression/resolution loss
ratio than simply using a low-pass (blur) filter on a single scan.
Marko Cebokli
-- Source code, list archive, and docs: http://www.mostang.com/sane/ To unsubscribe: echo unsubscribe sane-devel | mail majordomo@mostang.com
This archive was generated by hypermail 2b29 : Sun Dec 10 2000 - 04:05:04 PST