Jeff wrote:
>>
That's surprising! I thought several of the digital cameras used low pass
filters over their
sensors to eliminate aliasing? (Maybe I'm remembering scanners instead)
<<
Only the really high end digital bodies have offered anti-aliuasing filters.
The reason is, the
cost of the filters approaches that of the sensor chip. (According to one of
Foveon's Chief
Scientists). The E1 is one of the few "cheaper" cameras to include an
anti-alias filter. Even with
the anti-alias filter, reviewers have said that moire, which is partly related
to aliasing is
worse with the E1.
The discussion by CH. et al of whether to provide lenses with higher resolution
than the
anti-alias filter, is an issue of whether you want to be able to offer higher
resolution
bodies/sensors with corresponding changed anti-alias filter, in the future.
From a marketing point of view most users don't understand aliasing, so many
users remove the
anti-alias filter to "improve performance" anyway. Unfortunately it only
improves performance on
a certain type of image where you have certain expectations. On normal pictures
removing the
filter just adds noise to fine detailed areas, if they are at critical focus.
So there is not a
lot of upside to including an anti-alias filter, even though it is the right
thing to do. An
interesting aside, if the camera electronics do not limit (i.e. low noise),
then images produced
with an anti-alias filter can actualy be compressed more for the same image
quality.
The SD9 with Foveon chip, somebody mentioned, does not include an anti-alias
filter, but in
general it has a factor of 2 to 3 times lower aliasing (depending on color)
than a normal sensor.
The comparison gets sticky depending on how you want to do pixel counts when
comparing snsors with
different technology. In aliasing tests you need to compare both a B&W target
and color resolution
targets because the normal CCD's have much worse *color* aliasing and *color*
moire than Foveon.
The Nyquist frequency is much lower for R an B than for G for "normal" Beyer
sensors. The
interpolation software used in the normal sensors can make aliasing look worse
or look better,
depending both on the test target and the interp algorithm. This is probably
why there are often
contradictory test results. Also, if there is just a little camera shake, this
makes a perfect
anti-alias filter! I always thought Cannon should use it's IS lenses, in an
anti-alias mode, to
blur the lenses a little for anti-alias.
A point most people miss, with most digi-sensors, the resolution/aliasing is
dependent on how the
image is oriented relative the sensor. The resolution/Nyquist frequency is
~400wer on the
diagonal of the sensor, unless you use hexagonal pixels, in which case you need
a lot more for the
same coverage.
Regards,
Tim Hughes
< 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 >
|