Is the AA filter on the endangered list too?

In article
,
RichA wrote:

Seems like lighter and lighter AA filters are appearing in certain
cameras. The Leica has none. Could it be that this second-last
barrier to resolution (the Bayer filter being the last) isn't long for
the world?

only if someone can prove shannon/nyquist wrong, nor is it a barrier to
resolution.

In article , nospam
writes
In article
,
RichA wrote:

Seems like lighter and lighter AA filters are appearing in certain
cameras. The Leica has none. Could it be that this second-last
barrier to resolution (the Bayer filter being the last) isn't long for
the world?

only if someone can prove shannon/nyquist wrong,

Not necessarily - if the pixels are small enough to meet the Nyquist
sampling criteria of the optical image then the AA filter is
unnecessary. Since "never" is an extremely long time, this direct
implication of shannon/nyquist will occur long before anyone proves them
wrong.

nor is it a barrier to
resolution.

The optical AA filter is a barrier to resolution because it is an
imperfect AA filter according to shannon/nyquist. It is not flat in the
pass band and its blocking region is non-zero and the transition between
the two regions is not infinitely steep. The ratio of alias to signal
attenuation is a judgement call by the designers - none is perfect and
all are barriers to useful resolution to a greater or lesser degree.

Extremely small pixels combined with appropriate in-camera or on-chip
downsampling, similar to what is currently implemented in
one-dimensional audio systems, would allow a near perfect AA filter to
be implemented digitally, as well as eliminating the Bayer resolution
restriction.

Things are moving in that direction already with several cameras
offering sRAW options to output lower resolution images than the native
sensor format, with improved frame rate and/or "pixel" SNR.

It will be some time off, but optical AA filters will disappear
eventually.
--
Kennedy
Yes, Socrates himself is particularly missed;
A lovely little thinker, but a bugger when he's ****ed.
Python Philosophers (replace 'nospam' with 'kennedym' when replying)

In article , Kennedy McEwen
wrote:

Not necessarily - if the pixels are small enough to meet the Nyquist
sampling criteria of the optical image then the AA filter is
unnecessary. Since "never" is an extremely long time, this direct
implication of shannon/nyquist will occur long before anyone proves them
wrong.

that just makes the lens the anti-alias filter.

On 10-03-27 10:34 , Kennedy McEwen wrote:
In article , nospam
writes
In article
,
RichA wrote:

Seems like lighter and lighter AA filters are appearing in certain
cameras. The Leica has none. Could it be that this second-last
barrier to resolution (the Bayer filter being the last) isn't long for
the world?

only if someone can prove shannon/nyquist wrong,

Not necessarily - if the pixels are small enough to meet the Nyquist
sampling criteria of the optical image then the AA filter is
unnecessary. Since "never" is an extremely long time, this direct
implication of shannon/nyquist will occur long before anyone proves them
wrong.

nor is it a barrier to
resolution.

The optical AA filter is a barrier to resolution because it is an
imperfect AA filter according to shannon/nyquist. It is not flat in the
pass band and its blocking region is non-zero and the transition between
the two regions is not infinitely steep. The ratio of alias to signal
attenuation is a judgement call by the designers - none is perfect and
all are barriers to useful resolution to a greater or lesser degree.

This confirms what I said to nospam a couple days ago about how sharply
the filter is defined.


Extremely small pixels combined with appropriate in-camera or on-chip
downsampling, similar to what is currently implemented in
one-dimensional audio systems, would allow a near perfect AA filter to
be implemented digitally, as well as eliminating the Bayer resolution
restriction.

Things are moving in that direction already with several cameras
offering sRAW options to output lower resolution images than the native
sensor format, with improved frame rate and/or "pixel" SNR.

It will be some time off, but optical AA filters will disappear eventually.

I doubt it. No matter how fine the resolution of the sensor there is
always scene detail that falls at a line freq that will cause aliasing.


--
gmail originated posts are filtered due to spam.

In article , nospam
writes
In article , Kennedy McEwen
wrote:

Not necessarily - if the pixels are small enough to meet the Nyquist
sampling criteria of the optical image then the AA filter is
unnecessary. Since "never" is an extremely long time, this direct
implication of shannon/nyquist will occur long before anyone proves them
wrong.

that just makes the lens the anti-alias filter.

Exactly - no need for any additional optical AA filter.
--
Kennedy
Yes, Socrates himself is particularly missed;
A lovely little thinker, but a bugger when he's ****ed.
Python Philosophers (replace 'nospam' with 'kennedym' when replying)

In article , Alan Browne
writes
On 10-03-27 10:34 , Kennedy McEwen wrote:

Extremely small pixels combined with appropriate in-camera or on-chip
downsampling, similar to what is currently implemented in
one-dimensional audio systems, would allow a near perfect AA filter to
be implemented digitally, as well as eliminating the Bayer resolution
restriction.

Things are moving in that direction already with several cameras
offering sRAW options to output lower resolution images than the native
sensor format, with improved frame rate and/or "pixel" SNR.

It will be some time off, but optical AA filters will disappear eventually.

I doubt it. No matter how fine the resolution of the sensor there is
always scene detail that falls at a line freq that will cause aliasing.

No lens, not even a theoretical one, has infinite resolution.

Diffraction determines the maximum spatial resolution of a perfect lens.
That is a function of the relative aperture of the lens and the
wavelength of the light it is resolving. The former is limited by the
finite size of the lens aperture (be aware that the photographer's
approximation of relative aperture is increasingly in error at low f/#s
due to Lambert's Law), whilst millions of years of human evolution, the
temperature of the sun's corona and the transmission spectrum of the
earth's atmosphere place serious restrictions on the wavelength of light
your lens has to resolve. For an extremely good practical lens the
limit of resolution is much lower.

There is always a spatial frequency above which the lens MTF is zero,
and a sensor with pixels small enough to place the Nyquist above that
will never suffer from alias.
--
Kennedy
Yes, Socrates himself is particularly missed;
A lovely little thinker, but a bugger when he's ****ed.
Python Philosophers (replace 'nospam' with 'kennedym' when replying)

"Kennedy McEwen" wrote in message
...
In article , nospam
writes
In article , Kennedy McEwen
wrote:

Not necessarily - if the pixels are small enough to meet the Nyquist
sampling criteria of the optical image then the AA filter is
unnecessary. Since "never" is an extremely long time, this direct
implication of shannon/nyquist will occur long before anyone proves them
wrong.

that just makes the lens the anti-alias filter.

Exactly - no need for any additional optical AA filter.

But the lens is a really lousy AA filter. When diffraction isn't a problem,
the MTF has a very long tail. So you need many times your target resolution,
and then that number squared, data points.

And does post-demosaicing pixel bining actually work? Don't you lose DR
compared to larger pixels?

--
David J. Littleboy
Tokyo, Japan

In article , Kennedy McEwen
wrote:

that just makes the lens the anti-alias filter.

Exactly - no need for any additional optical AA filter.

not an additional one, but there is still an anti-alias filter in the
system.

On Sat, 27 Mar 2010 11:57:59 -0400, Alan Browne
wrote:


I doubt it. No matter how fine the resolution of the sensor there is
always scene detail that falls at a line freq that will cause aliasing.


The same can be said for your printer's pseudo-random dithering patterns,
or your LCD display, or phosphor dots and scan-lines on an
industry-standard CRT. This is why there are software solutions whenever it
is encountered. Unless you have the original scene and subjects with which
to compare the resulting display then no one's the wiser. There will never
be an exact 1:1 correlation between any subjects captured in any photograph
and reality. It's all approximations. One more added whenever you need to
get rid of aliasing artifacts. Be it by using misty atmospheric conditions,
haze-filters, soft-focus methods, in-camera sensor antialiasing filters, or
post-processing gaussian-blur digital manipulation.

Don't any of you people actually take photographs and do something with
them? If you ever had you'd know that you're now just mentally masturbating
about all of this.

In article , James Nagler
wrote:

The same can be said for your printer's pseudo-random dithering patterns,
or your LCD display, or phosphor dots and scan-lines on an
industry-standard CRT.

nonsense.

This is why there are software solutions whenever it
is encountered.

impossible to do post-capture.

Don't any of you people actually take photographs and do something with
them? If you ever had you'd know that you're now just mentally masturbating
about all of this.

as opposed to you physically masturbating?