Forgot Password
Pentax Camera Forums Home
 

Reply
Show Printable Version Search this Thread
09-24-2012, 09:35 AM   #1
Veteran Member




Join Date: Dec 2008
Location: Iowa
Photos: Gallery | Albums
Posts: 1,369
Current Cameras with No AA Filters (a running thread)

I've seen a lot of discussion going on about anti aliasing filters and what they would mean for the K-5 IIs since cameras without AA filters are new to Pentax shooters. I wanted to start a thread featuring cameras with no or weak AA filter examples so people could then go and look for themselves at those cameras to see if it appealed to them. I will start it off with the cameras I know of and then we can add more as others suggest it in future posts.

I will say from looking at the no AA filter list that all of them are cameras that I have read very positive things about their image quality, especially image resolution. This could bode well for the K-5 IIs.

No AA Filters

Leica M9 and M8
Nikon D800E
Ricoh GXR M mount Module (I believe the A12 28 and 50 mm units also have either no AA filter or weak ones)
Fuji X1-Pro (unique sensor design allows them to avoid moire)
Pentax 645D
Sigma Foveon cameras
Medium format backs

Weak AA Filter (I'm not as knowledgeable about this one, just did a quick Internet scan)

Olympus E5
Olympus E-1
Canon 1Ds
Canon 5D
Fuji X100


Last edited by Urkeldaedalus; 09-24-2012 at 10:07 AM.
09-24-2012, 09:43 AM   #2
Veteran Member




Join Date: Sep 2009
Location: Oregon
Photos: Gallery | Albums
Posts: 5,079
Also with no AA filters.

Fuji X100
All Sigma Foveon cameras
Pentax 645D (I believe)
All Medium Format digital backs.
09-24-2012, 10:06 AM   #3
Veteran Member




Join Date: Dec 2008
Location: Iowa
Photos: Gallery | Albums
Posts: 1,369
Original Poster
QuoteOriginally posted by Swift1 Quote
Also with no AA filters.

Fuji X100
All Sigma Foveon cameras
Pentax 645D (I believe)
All Medium Format digital backs.
Updated the list. I saw conflicting info on the Fuji X100, so I put it under the weak AA filter list for now.
09-24-2012, 10:08 AM   #4
Veteran Member
JinDesu's Avatar

Join Date: Jun 2011
Location: New York City
Photos: Gallery
Posts: 5,638
K-01 technically has a weaker filter than the k-5. Supposedly something like 1 filter vs 3.

09-24-2012, 10:14 AM   #5
Veteran Member
rparmar's Avatar

Join Date: Jan 2008
Photos: Gallery
Posts: 8,795
"Weak" is relative and maybe doesn't deserve to be enumerated. There's a big difference between a weak filter and no filter.
09-24-2012, 12:29 PM   #6
Veteran Member




Join Date: Dec 2008
Location: Iowa
Photos: Gallery | Albums
Posts: 1,369
Original Poster
QuoteOriginally posted by rparmar Quote
"Weak" is relative and maybe doesn't deserve to be enumerated. There's a big difference between a weak filter and no filter.
Good point. A "weak" AA filter is a bit harder to designate than a camera having no filter at all.

However, I just wanted to have another option to explain good results from cameras like the Olympus E5, which is said to produce excellent, sharp images with a supposedly weak AA filter still present.
09-25-2012, 09:33 AM   #7
Veteran Member
rparmar's Avatar

Join Date: Jan 2008
Photos: Gallery
Posts: 8,795
QuoteOriginally posted by Urkeldaedalus Quote
Good point. A "weak" AA filter is a bit harder to designate than a camera having no filter at all.

However, I just wanted to have another option to explain good results from cameras like the Olympus E5, which is said to produce excellent, sharp images with a supposedly weak AA filter still present.
Thing is, you don't know if that is a matter of the sensor, the image processing pipeline, or even the lens used in the tests. It's the internet. People write a lot of stuff with no facts behind it.

09-25-2012, 06:43 PM   #8
Veteran Member




Join Date: Jun 2009
Location: Chicago Suburbs.
Photos: Gallery
Posts: 501
Supposedly, from what I read on the internet (beware), that higher and higher megapixels negate the need for any kind of Anti-Aliasing filter because the frequency of any moire-provoking pattern is too small to ever see. Makes sense for a medium format sensor to not need an AA filter, if the premise is true. ?????
09-27-2012, 03:22 AM   #9
Veteran Member
kh1234567890's Avatar

Join Date: Sep 2010
Location: Manchester, UK
Photos: Gallery
Posts: 2,653
The culprit is the Bayer matrix. If you insist on viewing the world through a tea strainer you'll get subsampling artefacts.

Making the sensor pixels smaller just introduces other undesirable side effects such as diffraction, noise etc. The AA filter is more than something that just makes the image fuzzy in all directions, if properly designed it would be matched to the sensor pixel spacing and to the software demosaicing algorithm.
09-28-2012, 12:53 AM   #10
Veteran Member




Join Date: Jan 2010
Location: Southern England
Posts: 623
QuoteOriginally posted by kh1234567890 Quote
The culprit is the Bayer matrix. If you insist on viewing the world through a tea strainer you'll get subsampling artefacts.

Making the sensor pixels smaller just introduces other undesirable side effects such as diffraction, noise etc. The AA filter is more than something that just makes the image fuzzy in all directions, if properly designed it would be matched to the sensor pixel spacing and to the software demosaicing algorithm.
I have to disagree about smaller pixels (i.e. bigger pixel-counts). Assuming the same end pixel-count for the image (i.e. when it's displayed/printed), noise and the visibility of lens limitations will be the same, regardless of how large you make the pixel-count at the time of capture. The only thing wrong with large pixel-counts are the practical consequences (i.e. time/storage and heat, probably). I can see a continuing trend of larger and larger pixel-counts, as they offer the only proper way of reducing the need for strong anti-aliasing. (There's an analogy here to digital audio, by the way.)

I also disagree that the "culprit" is the Bayer matrix. The alternative would be a Foveon sensor. If we accept that the main limiting factor for pixel-count is the practicality issue (time, storage etc.), we'll note that a Foveon sensor presents 3x the data of a Bayer sensor (for a given pixel-count). This means that the Bayer sensor could have 3x the pixel-count for similar data-handling requirements, and that would reduce the need for anti-aliasing strength to that of the Foveon sensor (or at least similar to).
09-28-2012, 02:14 AM   #11
Veteran Member
kh1234567890's Avatar

Join Date: Sep 2010
Location: Manchester, UK
Photos: Gallery
Posts: 2,653
QuoteOriginally posted by m42man Quote
I also disagree that the "culprit" is the Bayer matrix. The alternative would be a Foveon sensor. If we accept that the main limiting factor for pixel-count is the practicality issue (time, storage etc.), we'll note that a Foveon sensor presents 3x the data of a Bayer sensor (for a given pixel-count). This means that the Bayer sensor could have 3x the pixel-count for similar data-handling requirements, and that would reduce the need for anti-aliasing strength to that of the Foveon sensor (or at least similar to).
The problem is subsampling - in the Bayer sensor there are physical gaps between elements of the same colour sensitivity, this is what leads to Moire artefacts. You can oversample - have more of smaller elements and then low pass filter in software or you can use a birefringent physical AA filter to spread the light just enough in the horizontal and vertical directions only so that a point source image always covers one or more of the rectangularly arranged sensing elements. Smaller sensing elements also collect fewer photons per exposure giving more quantisation noise. The Foveon sensor stacks the three colour sensing elements on top of each other (like a colour film would), making the inter-element gaps much smaller.

Last edited by kh1234567890; 09-28-2012 at 02:33 AM.
09-28-2012, 11:25 PM   #12
Veteran Member




Join Date: Jan 2010
Location: Southern England
Posts: 623
QuoteOriginally posted by kh1234567890 Quote
The problem is subsampling - in the Bayer sensor there are physical gaps between elements of the same colour sensitivity, this is what leads to Moire artefacts. You can oversample - have more of smaller elements and then low pass filter in software or you can use a birefringent physical AA filter to spread the light just enough in the horizontal and vertical directions only so that a point source image always covers one or more of the rectangularly arranged sensing elements. Smaller sensing elements also collect fewer photons per exposure giving more quantisation noise. The Foveon sensor stacks the three colour sensing elements on top of each other (like a colour film would), making the inter-element gaps much smaller.
Yes, my whole point is that the Bayer matrix will sample at a frequency similar to the Foveon if we use equal data rates as the basis for comparison (as opposed to equal pixel-counts). The Foveon pixel embraces all 3 colours vertically, the Bayer spreads them laterally, so (crudely, the reality is more complex) a 3-colour Bayer "superpixel" would need to have 3 pixels occupying the same space as a Foveon pixel - hence my suggestion that the Bayer sensor has 3x the pixel-count. Obviously, if you use equal pixel-counts as the basis for comparison, the Bayer would need stronger anti-aliasing as it's sampling at a lower frequency. (I do notice that the Sigma SD1 Foveon marketing describes the pixel-count as 48MP, but this refers to 16MP-worth of 3-colour pixels).

Regarding noise, if you print at the same resolution as the sensor, then, yes, you'll get more noise as you increase pixel-count. However, if you were to compare high and low pixel-count sensors when printing at a constant resolution (i.e. constant print size), you wouldn't see any difference. How can this be? Well (crudely) the printer will be combining ("binning") pixels in software to make bigger "virtual pixels", which have the same reduced quantisation noise as real bigger pixels. There's nothing wrong with smaller pixels - it's just the high data rates which are currently problematic.

Last edited by m42man; 09-28-2012 at 11:33 PM.
09-29-2012, 12:53 AM   #13
Veteran Member
kh1234567890's Avatar

Join Date: Sep 2010
Location: Manchester, UK
Photos: Gallery
Posts: 2,653
It is not pixel count or data rates but spatial subsampling - there are effectively physical gaps on the Bayer matrix where it is 'blind'. This is what gives Moire. Consider for example (however unlikely in real life) a situation where you have an image of a fine repeating rectangular pattern in blue, aligned so that it only falls on the red or green sensitive elements. Without the AA filter the Bayer sensor will not see it.
09-29-2012, 09:44 AM   #14
Inactive Account




Join Date: Dec 2008
Location: Ames, Iowa, USA
Photos: Albums
Posts: 2,965
Are there any tabulated comparisons of AA filter strengths (like gaussian, sinc equivalent, etc) for various Pentax cameras?

Conversely, how might one test for a rough measure?

Like: I have a K-x camera - can I expect edge blur at best to be like that resulting from a Gaussian(1) filter, Gaussian(2) filter, ....?

Dave in Iowa
09-29-2012, 02:55 PM   #15
Veteran Member
kh1234567890's Avatar

Join Date: Sep 2010
Location: Manchester, UK
Photos: Gallery
Posts: 2,653
A birefringent filter does not blur the image, it doubles it.

Reply

Bookmarks
  • Submit Thread to Facebook Facebook
  • Submit Thread to Twitter Twitter
  • Submit Thread to Digg Digg
Tags - Make this thread easier to find by adding keywords to it!
camera, cameras, canon, dslr, filter, filters, fuji, image, k-5, olympus, pentax, photography, thread
Thread Tools Search this Thread
Search this Thread:

Advanced Search


Similar Threads
Thread Thread Starter Forum Replies Last Post
Donation/information request - my school is running out of cameras! Steve Beswick Film SLRs and Compact Film Cameras 45 03-20-2014 08:43 PM
Running thread of consolidated Tea Party-related stupidity deadwolfbones General Talk 289 11-04-2011 08:17 PM
Running Thread of Catharsis shooz General Talk 54 05-24-2011 03:30 PM



All times are GMT -7. The time now is 03:56 AM. | See also: NikonForums.com, CanonForums.com part of our network of photo forums!
  • Red (Default)
  • Green
  • Gray
  • Dark
  • Dark Yellow
  • Dark Blue
  • Old Red
  • Old Green
  • Old Gray
  • Dial-Up Style
Hello! It's great to see you back on the forum! Have you considered joining the community?
register
Creating a FREE ACCOUNT takes under a minute, removes ads, and lets you post! [Dismiss]
Top