K-5 IIs -- The Pros and Cons of Omitting an AA-Filter

audiobomber · 10-25-2012, 02:57 PM

Originally posted by Ecaterin

I, too, will keep chiming in on these "no AA filter OMG!" threads along with the other users of non-AA filter cameras and say that under real world shooting conditions (portraiture, landscape, you name it) I have yet to see moire on anything shot with my Fuji X100.

The X100 has an AA filter. This Luminous Landscape article contains a list of all the FF and smaller bodies with no AA:
Ricoh Mount A12, Sigma SD1/SD1 Merrill, Sigma DP1 Merrill, Sigma DP2 Merrill, Fuji X-Pro 1 and the Nikon D800E.
The Naked Sensor

Written before the release of the Pentax of course, but well after the release of the X100, which is discussed in the article.

Parallax · 10-25-2012, 03:10 PM

The X-100 has a different array that does not require a filter. FUJIFILM New Zealand

Ecaterin · 10-25-2012, 03:11 PM

Originally posted by audiobomber

The X100 has an AA filter. This Luminous Landscape article contains a list of all the FF and smaller bodies with no AA:
Ricoh Mount A12, Sigma SD1/SD1 Merrill, Sigma DP1 Merrill, Sigma DP2 Merrill, Fuji X-Pro 1 and the Nikon D800E.
The Naked Sensor

Written before the release of the Pentax of course, but well after the release of the X100, which is discussed in the article.

*See Ecat zip out into the wilds of Google and read a bunch of stuff right quick!*

Well, this is news to me! When I did my initial research and purchase of the X100, word on the street was 'no AA filter.' Looks like that's been amended to 'very weak AA filter' - though the Big Name reviews still read as 'none' and have not been changed. So now I'm wondering, does the X-Pro-1 have a weak AA filter, or none? Reading around, the consensus seems to be 'none.' The randomizing of the sensor tech and the use of clever algorithms is still damn impressive for this as an emerging technology - it's clearly The Way Of The Future.

Ecaterin · 10-25-2012, 03:12 PM

Originally posted by Parallax

The X-100 has a different array that does not require a filter. FUJIFILM New Zealand

LOL - man, I love PF - you can't even keep up with reading awesome links fast enough to correct your own posts LOL!

audiobomber · 10-25-2012, 03:19 PM

Originally posted by Parallax

The X-100 has a different array that does not require a filter. FUJIFILM New Zealand

No, that article is about the X-Pro1 (X-Trans) sensor. The only mention of the X100 is in relation to APS-C:

"As the flagship model of the X series, the X-Pro1 has the mission of delivering high picture quality that meets the demands of professional photographers and offering the size that provides advanced maneuverability at the same time. To fulfill both of these major objectives, APS was the optimum format, as proven by X100. ・X-Pro1 has gone one step further to achieve the level of picture quality that outperforms that of other interchangeable lens cameras and even full-size SLRs."

The step further is the X-Trans sensor with a non-Bayer array.

kh1234567890 · 10-25-2012, 03:28 PM

Originally posted by Ecaterin

Looks like that's been amended to 'very weak AA filter' - though the Big Name reviews still read as 'none' and have not been changed.

This 'weak AA filter' feature has always puzzled me. How is it done, what is the point ?

Perhaps someone who's dabbled in Physics or Crystallography more recently than I have would care to explain ?

falconeye · 10-25-2012, 04:35 PM

Originally posted by kh1234567890

This 'weak AA filter' feature has always puzzled me. How is it done, what is the point ?

Perhaps someone who's dabbled in Physics or Crystallography more recently than I have would care to explain ?

Crystallography is physics, i.e. solid state physics.

What Class A decided to skip in his great opening post is this:

The so-called AA filter is no anti-alias filter. It is only that everybody (falsely) calls it so.

Early medium format digital cameras with their small fill factor due to lack of a micro lens array had aliasing artefacts: staircase edges, spurious resolution and luminance moiré. No current camera anymore has this problem, they all have AA fiters curing the problem aka micro lens array. Therefore, it is technically wrong to say the Leica Monochrome has no AA filter: it has and is (or should be) unable to produce a moiré pattern.

Here, we actually talk about the false color problem: false color specular highlights (overlooked in this thread so far) and false color moiré. The cure is a pair of birefringent crystals (which I always call Bayer-AA filter rather than AA filter if you look up my posts): it ONLY works in combination with the micro lens array and TOGETHER they change the Point Spread Function (PSF) from a point into a square with the length of 2 pixels (such that each ray always hits all 4 RGGB Bayer pixels which is required to record the ray's color without any interpolation).

This is for birefringent crystals which distort the anormal ray by X times the pixel pitch where X=1. In general, the PSF edge length becomes 1+X times the pixel pitch.

A weak Bayer-AA filter is obtained by making the crystal thinner, leading to X<1. At X=0, the Bayer-AA filter is absent.

I tried to measure X for the D800. AFAIK, it is the only attempt at this so far. I obtained X=~0.8 (80%). 0.8 may seem to be close to 1. But values near 1 damp the MTF near Nyquist significantly and going from 1 to 0.8 is a significant change.

As a result, the D800 has a so-called weak Bayer-AA filter which provides sharp images with a small remaining risk for false color moiré.

In order to be able to recommend for or against the K-5IIs for a given application, one would have to know X for the K-5II. Unfortunately, vendors don't provide this information. This may be ok in general, but where there is a choice (D800, K-5II), it is an intolerable lack of information (IMHO) the buyer should insist to obtain.

Last edited by falconeye; 10-25-2012 at 05:14 PM.

falconeye · 10-25-2012, 04:46 PM - **1 Like**

BTW

I am not convinced that the Fuji X-Pro1 sensor is a good solution. It solves a certain class of false color moré patterns. But while it minimizes the risk at one pattern frequency, it increases the possibility at a few other frequencies, and it does NOT solve the false color specular highlight problem (a problem I see as aften as false color moiré, typically in hair or fur lit by sun shine or flash).

Moreover, it seems to reduce the maximum resolution in the luminance channel a bit, compared to the original Bayer layout which reaches the Nyqust frequency. Which may be a reason why e.g. Adobe as trouble to support it. Theoretically, the Fuji sensor shouldn't be worse (and for a pattern known to be B&W t isn't). But the required color interpolation in the general case is less local. The points where all 3 colors are adjacent are now sparse.

So, in a direct comparison with a combination of weak Bayer-AA filter and a classic Bayer layout, I don't know yet which approach would win.

audiobomber · 10-25-2012, 04:54 PM

Originally posted by falconeye

So, in a direct comparison with a combination of weak Bayer-AA filter and a classic Bayer layout, I don't know yet which approach would win.

I'm not sure what you mean by "win". Are you talking about moire control, resolution, general performance, something else?

Winder · 10-25-2012, 05:02 PM

Originally posted by audiobomber

These are examples of nature causing moire alright, but the parrot feathers were sharpened to enhance the moire. The fly photo needed to be stopped down a lot more for reasonable DOF.

This is true, but these were also taken with a camera that had an AA filter. The results would be more significant if a camera without an AA filter had been used.

kh1234567890 · 10-25-2012, 05:03 PM

Originally posted by falconeye

Crystallography is physics, i.e. solid state physics.

You try telling that to metallurgists ...

Originally posted by falconeye

A weak Bayer-AA filter is obtained by making the crystal thinner, leading to X<1. At X=0, the Bayer-AA filter is absent.
I tried to measure X for the D800. AFAIK, it is the only attempt at this so far. I obtained X=~0.8 (80%).
As a result, the D800 has a so-called weak Bayer-AA filter which provides sharp images with a small remaining risk for false color moiré.

Doesn't making the spread between the ordinary and extraordinary rays smaller (X<1) sort of defeat the whole point of having the filter ? Surely in that case you can't be sure of covering a full sensel quad every time, making thing worse ? Wouldn't it make more sense to use a different cut of the birefringent crystal as to make the secondary images weaker ?

falconeye · 10-25-2012, 05:32 PM

Originally posted by kh1234567890

use a different cut of the birefringent crystal as to make the secondary images weaker ?

You wouldn't be able to tell the primary and secondary rays apart once captured. To make the crystal thinner is a nice way to interpolate between Bayer-AA and no Bayer-AA filter.

Even w/o any Bayer-AA filter, the demosaicing will in general reconstruct the right color, e.g., for a single edge, by making use of a set of most probable assumptions. A weak Bayer-AA filter helps to not break said assumptions.

Since you asked, I really wanted to explain what the Bayer-AA filter does. I fear that I am not the expert and this isn't the right forum to discuss alternatives.

**Digitalis** · 10-25-2012, 05:44 PM

Originally posted by falconeye

Since you asked, I really wanted to explain what the Bayer-AA filter does. I fear that I am not the expert and this isn't the right forum to discuss alternatives.

That's perhaps the biggest problem of all - none of us are experts. Most of what we have to go on is, experience, conjecture and information gleaned from published sources that are hard to corroborate or verify. As I recall Hasselblad did quite a bit or research into digital imaging* decades before it became mainstream - and I recall reading a technical publication in the 1990's on their research into digital backs and all the problems they had to eliminate to produce acceptable image quality from it....I'll have to find that.

And Falk ,you are correct: the Leica Monochrom is immuneto luminance moire. I have also had some interesting success in using red,green,blue coloured flash filters to produce colour images from the Monochrom, I have also experimented with cyan,magenta, yellow with an unfiltered Luminance image for the K channel, It turned out quite well. With a few tweaks it would me entirely possible for me to turn out CMYK images for press printing without going through the rather torturous RGB>CMYK conversion process.

*Kodak was unsurprisingly late to the game, however there should be a few technical publications on their sensors floating around. But from what I have read the Bayer AA filter as you describe it is not a part of the sensor package, like the micro-lenses are - it is installed by the camera manufacturer.

Dr_who · 10-25-2012, 07:57 PM

Originally posted by JinDesu

So what I mean is - these pictures that we see online that aren't full resolution: couldn't the moire just be a result of the downsizing algorithm? For my example image:

When you click the image for full resolution (well, larger anyways) - the moire is gone. But because of how the image is reduced and interpreted, the moire is shown. I think we may have to be a little picky when selecting images that represent moire.

The humorous part is that I walked by a bunch of window screens left on the street yesterday, and with my naked eye looking at them from an angle, it looked like patterns galore. I do not doubt the camera would capture the same.

Actually when I enlarge it the moire gets quite a bit worse....at least on my work computers monitor. At home moire on the small image is barely there, and enlarged its perfect. But I totally get what your saying, as your also introducing a another variable in what can cause moire. The pixel resolution of ones monitor.

JinDesu · 10-25-2012, 07:59 PM

Originally posted by Dr_who

Actually when I enlarge it the moire gets quite a bit worse....at least on my work computers monitor. At home moire on the small image is barely there, and enlarged its perfect. But I totally get what your saying, as your also introducing a another variable in what can cause moire. The pixel resolution of ones monitor.

Yeah - I never thought about it until this thread. For example, the bug eye in one of the pictures listed above. I highly suspect that at full resolution, there isn't moire. It's the downsizing and viewing medium.

10-25-2012, 03:19 PM	#50
audiobomber Veteran Member Join Date: May 2008 Location: Sudbury, Ontario Photos: Albums Posts: 6,806	Originally posted by Parallax The X-100 has a different array that does not require a filter. FUJIFILM New Zealand No, that article is about the X-Pro1 (X-Trans) sensor. The only mention of the X100 is in relation to APS-C: "As the flagship model of the X series, the X-Pro1 has the mission of delivering high picture quality that meets the demands of professional photographers and offering the size that provides advanced maneuverability at the same time. To fulfill both of these major objectives, APS was the optimum format, as proven by X100. ・X-Pro1 has gone one step further to achieve the level of picture quality that outperforms that of other interchangeable lens cameras and even full-size SLRs." The step further is the X-Trans sensor with a non-Bayer array. Last edited by audiobomber; 10-25-2012 at 03:26 PM.

10-25-2012, 04:46 PM - 1 Like	#53
falconeye Veteran Member Join Date: Jan 2008 Location: Munich, Alps, Germany Photos: Gallery Posts: 6,871	BTW I am not convinced that the Fuji X-Pro1 sensor is a good solution. It solves a certain class of false color moré patterns. But while it minimizes the risk at one pattern frequency, it increases the possibility at a few other frequencies, and it does NOT solve the false color specular highlight problem (a problem I see as aften as false color moiré, typically in hair or fur lit by sun shine or flash). Moreover, it seems to reduce the maximum resolution in the luminance channel a bit, compared to the original Bayer layout which reaches the Nyqust frequency. Which may be a reason why e.g. Adobe as trouble to support it. Theoretically, the Fuji sensor shouldn't be worse (and for a pattern known to be B&W t isn't). But the required color interpolation in the general case is less local. The points where all 3 colors are adjacent are now sparse. So, in a direct comparison with a combination of weak Bayer-AA filter and a classic Bayer layout, I don't know yet which approach would win. Last edited by falconeye; 10-25-2012 at 04:58 PM.

10-25-2012, 05:32 PM	#57
falconeye Veteran Member Join Date: Jan 2008 Location: Munich, Alps, Germany Photos: Gallery Posts: 6,871	Originally posted by kh1234567890 use a different cut of the birefringent crystal as to make the secondary images weaker ? You wouldn't be able to tell the primary and secondary rays apart once captured. To make the crystal thinner is a nice way to interpolate between Bayer-AA and no Bayer-AA filter. Even w/o any Bayer-AA filter, the demosaicing will in general reconstruct the right color, e.g., for a single edge, by making use of a set of most probable assumptions. A weak Bayer-AA filter helps to not break said assumptions. Since you asked, I really wanted to explain what the Bayer-AA filter does. I fear that I am not the expert and this isn't the right forum to discuss alternatives. Last edited by falconeye; 10-25-2012 at 05:40 PM.

10-25-2012, 05:44 PM	#58
Digitalis Site Supporter Join Date: Mar 2009 Location: Melbourne, Victoria Photos: Gallery Posts: 11,694	Originally posted by falconeye Since you asked, I really wanted to explain what the Bayer-AA filter does. I fear that I am not the expert and this isn't the right forum to discuss alternatives. That's perhaps the biggest problem of all - none of us are experts. Most of what we have to go on is, experience, conjecture and information gleaned from published sources that are hard to corroborate or verify. As I recall Hasselblad did quite a bit or research into digital imaging* decades before it became mainstream - and I recall reading a technical publication in the 1990's on their research into digital backs and all the problems they had to eliminate to produce acceptable image quality from it....I'll have to find that. And Falk ,you are correct: the Leica Monochrom is immuneto luminance moire. I have also had some interesting success in using red,green,blue coloured flash filters to produce colour images from the Monochrom, I have also experimented with cyan,magenta, yellow with an unfiltered Luminance image for the K channel, It turned out quite well. With a few tweaks it would me entirely possible for me to turn out CMYK images for press printing without going through the rather torturous RGB>CMYK conversion process. Kodak was unsurprisingly late to the game, however there should be a few technical publications on their sensors floating around. But from what I have read the Bayer AA filter as you describe it is not a part of the sensor package, like the micro-lenses are - it is installed by the camera manufacturer. Last edited by Digitalis; 10-25-2012 at 05:58 PM.*

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
From K20D to 645D, Pros and Cons.	Reportage	Pentax DSLR Discussion	12	11-10-2010 03:13 PM
Pentax K-5 pros/cons	psychdoc	Pentax K-5 & K-5 II	11	11-08-2010 12:41 PM
Decision: Pentax K10D, K20D, or K7? Pros and cons?	Lulerfly	Pentax DSLR Discussion	32	11-04-2010 09:14 AM
Pros and cons of the Pentax KM/2000	lesmore49	Pentax DSLR Discussion	109	05-01-2009 11:51 PM
Sigma 28-70 F/2.8 (Pros & Cons)	GLThorne	Pentax SLR Lens Discussion	8	03-24-2007 07:10 AM

10-25-2012, 02:57 PM	#46
audiobomber Veteran Member Join Date: May 2008 Location: Sudbury, Ontario Photos: Albums Posts: 6,806	Originally posted by Ecaterin I, too, will keep chiming in on these "no AA filter OMG!" threads along with the other users of non-AA filter cameras and say that under real world shooting conditions (portraiture, landscape, you name it) I have yet to see moire on anything shot with my Fuji X100. The X100 has an AA filter. This Luminous Landscape article contains a list of all the FF and smaller bodies with no AA: Ricoh Mount A12, Sigma SD1/SD1 Merrill, Sigma DP1 Merrill, Sigma DP2 Merrill, Fuji X-Pro 1 and the Nikon D800E. The Naked Sensor Written before the release of the Pentax of course, but well after the release of the X100, which is discussed in the article.

10-25-2012, 03:10 PM	#47
Parallax Senior Moderator Loyal Site Supporter Join Date: Apr 2008 Location: South Dakota Photos: Gallery Posts: 19,332	The X-100 has a different array that does not require a filter. FUJIFILM New Zealand

10-25-2012, 03:11 PM	#48
Ecaterin Veteran Member Join Date: Nov 2009 Location: Seattle Photos: Gallery \| Albums Posts: 723	Originally posted by audiobomber The X100 has an AA filter. This Luminous Landscape article contains a list of all the FF and smaller bodies with no AA: Ricoh Mount A12, Sigma SD1/SD1 Merrill, Sigma DP1 Merrill, Sigma DP2 Merrill, Fuji X-Pro 1 and the Nikon D800E. The Naked Sensor Written before the release of the Pentax of course, but well after the release of the X100, which is discussed in the article. See Ecat zip out into the wilds of Google and read a bunch of stuff right quick! Well, this is news to me! When I did my initial research and purchase of the X100, word on the street was 'no AA filter.' Looks like that's been amended to 'very weak AA filter' - though the Big Name reviews still read as 'none' and have not been changed. So now I'm wondering, does the X-Pro-1 have a weak AA filter, or none? Reading around, the consensus seems to be 'none.' The randomizing of the sensor tech and the use of clever algorithms is still damn impressive for this as an emerging technology - it's clearly The Way Of The Future.

10-25-2012, 03:12 PM	#49
Ecaterin Veteran Member Join Date: Nov 2009 Location: Seattle Photos: Gallery \| Albums Posts: 723	Originally posted by Parallax The X-100 has a different array that does not require a filter. FUJIFILM New Zealand LOL - man, I love PF - you can't even keep up with reading awesome links fast enough to correct your own posts LOL!

10-25-2012, 03:28 PM	#51
kh1234567890 Veteran Member Join Date: Sep 2010 Location: Manchester, UK Photos: Gallery Posts: 2,653	Originally posted by Ecaterin Looks like that's been amended to 'very weak AA filter' - though the Big Name reviews still read as 'none' and have not been changed. This 'weak AA filter' feature has always puzzled me. How is it done, what is the point ? Perhaps someone who's dabbled in Physics or Crystallography more recently than I have would care to explain ?

10-25-2012, 04:35 PM	#52
falconeye Veteran Member Join Date: Jan 2008 Location: Munich, Alps, Germany Photos: Gallery Posts: 6,871	Originally posted by kh1234567890 This 'weak AA filter' feature has always puzzled me. How is it done, what is the point ? Perhaps someone who's dabbled in Physics or Crystallography more recently than I have would care to explain ? Crystallography is physics, i.e. solid state physics. What Class A decided to skip in his great opening post is this: The so-called AA filter is no anti-alias filter. It is only that everybody (falsely) calls it so. Early medium format digital cameras with their small fill factor due to lack of a micro lens array had aliasing artefacts: staircase edges, spurious resolution and luminance moiré. No current camera anymore has this problem, they all have AA fiters curing the problem aka micro lens array. Therefore, it is technically wrong to say the Leica Monochrome has no AA filter: it has and is (or should be) unable to produce a moiré pattern. Here, we actually talk about the false color problem: false color specular highlights (overlooked in this thread so far) and false color moiré. The cure is a pair of birefringent crystals (which I always call Bayer-AA filter rather than AA filter if you look up my posts): it ONLY works in combination with the micro lens array and TOGETHER they change the Point Spread Function (PSF) from a point into a square with the length of 2 pixels (such that each ray always hits all 4 RGGB Bayer pixels which is required to record the ray's color without any interpolation). This is for birefringent crystals which distort the anormal ray by X times the pixel pitch where X=1. In general, the PSF edge length becomes 1+X times the pixel pitch. A weak Bayer-AA filter is obtained by making the crystal thinner, leading to X<1. At X=0, the Bayer-AA filter is absent. I tried to measure X for the D800. AFAIK, it is the only attempt at this so far. I obtained X=~0.8 (80%). 0.8 may seem to be close to 1. But values near 1 damp the MTF near Nyquist significantly and going from 1 to 0.8 is a significant change. As a result, the D800 has a so-called weak Bayer-AA filter which provides sharp images with a small remaining risk for false color moiré. In order to be able to recommend for or against the K-5IIs for a given application, one would have to know X for the K-5II. Unfortunately, vendors don't provide this information. This may be ok in general, but where there is a choice (D800, K-5II), it is an intolerable lack of information (IMHO) the buyer should insist to obtain. Last edited by falconeye; 10-25-2012 at 05:14 PM.

10-25-2012, 04:54 PM	#54
audiobomber Veteran Member Join Date: May 2008 Location: Sudbury, Ontario Photos: Albums Posts: 6,806	Originally posted by falconeye So, in a direct comparison with a combination of weak Bayer-AA filter and a classic Bayer layout, I don't know yet which approach would win. I'm not sure what you mean by "win". Are you talking about moire control, resolution, general performance, something else?

10-25-2012, 05:02 PM	#55
Winder Veteran Member Join Date: Apr 2010 Location: Tennessee Posts: 6,617	Originally posted by audiobomber These are examples of nature causing moire alright, but the parrot feathers were sharpened to enhance the moire. The fly photo needed to be stopped down a lot more for reasonable DOF. This is true, but these were also taken with a camera that had an AA filter. The results would be more significant if a camera without an AA filter had been used.

10-25-2012, 05:03 PM	#56
kh1234567890 Veteran Member Join Date: Sep 2010 Location: Manchester, UK Photos: Gallery Posts: 2,653	Originally posted by falconeye Crystallography is physics, i.e. solid state physics. You try telling that to metallurgists ... Originally posted by falconeye A weak Bayer-AA filter is obtained by making the crystal thinner, leading to X<1. At X=0, the Bayer-AA filter is absent. I tried to measure X for the D800. AFAIK, it is the only attempt at this so far. I obtained X=~0.8 (80%). As a result, the D800 has a so-called weak Bayer-AA filter which provides sharp images with a small remaining risk for false color moiré. Doesn't making the spread between the ordinary and extraordinary rays smaller (X<1) sort of defeat the whole point of having the filter ? Surely in that case you can't be sure of covering a full sensel quad every time, making thing worse ? Wouldn't it make more sense to use a different cut of the birefringent crystal as to make the secondary images weaker ?

10-25-2012, 07:57 PM	#59
Dr_who Veteran Member Join Date: Jun 2009 Location: Edmonton, Alberta, Canada Photos: Gallery \| Albums Posts: 777	Originally posted by JinDesu So what I mean is - these pictures that we see online that aren't full resolution: couldn't the moire just be a result of the downsizing algorithm? For my example image: When you click the image for full resolution (well, larger anyways) - the moire is gone. But because of how the image is reduced and interpreted, the moire is shown. I think we may have to be a little picky when selecting images that represent moire. The humorous part is that I walked by a bunch of window screens left on the street yesterday, and with my naked eye looking at them from an angle, it looked like patterns galore. I do not doubt the camera would capture the same. Actually when I enlarge it the moire gets quite a bit worse....at least on my work computers monitor. At home moire on the small image is barely there, and enlarged its perfect. But I totally get what your saying, as your also introducing a another variable in what can cause moire. The pixel resolution of ones monitor.

Thread Tools	Search this Thread
	Search this Thread: Advanced Search

10-25-2012, 07:59 PM	#60
JinDesu Veteran Member Join Date: Jun 2011 Location: New York City Photos: Gallery Posts: 5,638	Originally posted by Dr_who Actually when I enlarge it the moire gets quite a bit worse....at least on my work computers monitor. At home moire on the small image is barely there, and enlarged its perfect. But I totally get what your saying, as your also introducing a another variable in what can cause moire. The pixel resolution of ones monitor. Yeah - I never thought about it until this thread. For example, the bug eye in one of the pictures listed above. I highly suspect that at full resolution, there isn't moire. It's the downsizing and viewing medium.