Actually I believe some of the early Sony CD players advertised 8x
oversampling A/D converter. The idea being that with oversampling you
can use a lower resolution A/D, then filter the result to get CD level
accuracy. This is similar to taking several values and averaging to get
more resolution. The advertising "implied" that some how oversampled
means lower noise, but the technique is ususally used to get rid of
high frequency out of band noise and *aliasing* effects with a cheaper
analog filter on the outpue of an A/D converter.
Aliasing is similar to the lines-per-mm discussion and sampling in a
scanner. If you sample near the rate of the frequency of change, such
as density of light and dark lines in test film, you alias down to a
lower frequency and get all white, all dark or some mixture in between.
If the sampling rate is just slightly off, you get a low frequency line
that would appears as large bands of light and dark. This is aliasing
higher frequendy to a lower frequendy.
One of the interesting effects i have seen with film scanning is the
frequency of the grain in film and the sampling rate (pixel density) of
the scanner. The pixel aperture can be narrow relative to the pixel
sample rate. That is, the pixel sampling will magnify a small sample of
the grain into the full pixel. Multi-sampling on the LS-2000 can help
average out the sampling effects much like a smoothing (anti-alias)
filter does to audio. I have seen some 400 speed film give blotchy
looking effects to someone's face, while the regular printed photo from
the negative seems to average out the grain giving a much more pleasing
result.
This is another reason it is better to sample the original at the
highest pixel density and use software to down-sample the image to a
lower pixel density, it can average out the aliasing effects better
that way, by averaging more local pixels together.
Wayne
Hopkinton, MA
< This message was delivered via the Olympus Mailing List >
< For questions, mailto:owner-olympus@xxxxxxxxxxxxxxx >
< Web Page: http://Zuiko.sls.bc.ca/swright/olympuslist.html >
|