cbagwell@sprynet.com wrote:
> This brings up a topic that was discussed a while back. Should we
> have a new classification for scanners that have uneven scan
> resolutions? Those would be allowed to interpolate so that you got
> an even 600 x 600 dpi. I'm assuming the vertical 600 would be a
> better version than pure interpolation alone could produce?
Its hard to tell what is best. The scanner documentation I have looked at isn't
too clear about what exactly the number of bits of colour depth, or the
oversampled dpi figures actually mean - they like to quote big numbers for the
spec., and avoid justifying them.
If the scanners interpolate from an oversampled analogue signal from the CCD,
they could achieve substantially better oversampled resolution, in the
direction of scan, than a post processing interpolator on the host could do.
That is the direction which already has the natural higher resolution, though.
Across the width of the scan they can interpolate before gamma correction. I
think that would give a better result than interpolating on the host after
gamma correction. You cannot really undo the gamma correction on the host for
interpolation, as you have already lost considerable brightness resolution in
the scanner's gamma corrector. Generally the 30 or 36 bits quoted on the
scanner box is before gamma correction, and only 24 bits is passed back to the
host..
So, I think a 600x1200 scanner outputting a 1200x1200 image will give a better
result than interpolating to square pixels on the host. I also think that this
level of interpolation is quite meaningful. I really doubt than interpolating
much further has any purpose. Before I owned a scanner I thought high
resolution seemed a wonderful idea. Seeing the size of scan files quickly
changed my mind! The interpolated scans of 35mm film I have seen look pretty
dire.
Regards,
Steve
-- 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 : Wed Nov 29 2000 - 18:14:36 PST