> #1 Resize the scanned pixels so the image is 4 inches by 4
> inches within photoshop keeping the 2900 dots of data from the
> original scan. (I am not at all certain how this works, but this
> is what I got from reading "A Few Scanning Tips" by Wayne Fulton
> and Photoshop 6.0 seems to do it.)
>
Many (most or even all?) print drivers will take your 2900 / 8 = 363.5 dpi
image and perform a quick and dirty decimation to get down to the printers
200 dpi. By decimation I mean they will throw away extra pixels until the
resolution is 200 dpi. Likewise, if you feed the printer less than 200 dpi
then the print driver will simply duplicate adjacent pixels until it has 200
dpi.
Remember, when I say "200 dpi" it is only because that is the example
resolution you used. I will send another reply that has an Acrobat file I
created that tests this for your printer. For my printer (an Alps MD5000)
the magic resolution is 300 dpi. The other post may or may not pass through
the list because the size is 56 kB.
>
> #2 Resample in photoshop to "convert" the 2900 dots of data to
> 800 dots of data at 200 dpi which I send to the printer.
> >
>
> This should give the best results - the bi-cubic resampling used by
> Photoshop (make sure you have your preferences set to use bi-cubic by
> default) is near optimal. I am aware of only one program that uses the
> optimal, sinc, resampling - PanoTools at
> http://www.fh-furtwangen.de/~dersch/ (closed for the summer).
>
> >
> #3 Stick the slide into the scanner and get the "desired" 800 by
> 800 bits of data from the selected portion of the slide. i.e.
> somehow use the Twain drivers to get the resolution I want.
> >
>
Should be equivalent to #1 - decimation of pixels to get the right
resolution.
> >
> I recognize in the real world the choices will never be this
> clear, but can one generalize that resampling to a *lower*
> resolutions is better/worse then resizing under certain
> circumstances?
>
> It seems to me scanner software would not even offer the option
> of lower resolution scans if the quality were always better for
> resampling and resizing.
> >
>
But it is much easier to brag about scan time. Image quality is much harder
to measure.