Topic: Remove AA filter? (Read 23909 times)

markhout · « **on:** April 21, 2009, 08:00:28 pm »

Was curious about the effects of removing the AA filter to gain some extra perceived resolution and minimize capture sharpening. Came across this site: http://www.maxmax.com/nikon_d200hr.htm

Can anyone report on the practical pros and cons? I am shooting mainly landscapes, so I'm not overly worried about moire.

Thanks,

Mark

spotmeter · « **Reply #1 on:** April 21, 2009, 08:40:19 pm »

Quote from: markhout

Was curious about the effects of removing the AA filter to gain some extra perceived resolution and minimize capture sharpening. Came across this site:
http://www.maxmax.com/nikon_d300.htm

Can anyone report on the practical pros and cons? I am shooting mainly landscapes, so I'm not overly worried about moire.

Thanks,

Mark

I've been thinking about this conversion for my Canon 5D. As a landscape photographer, I am always looking to improve resolution.

I have always wondered why Canon and Nikon use an AA filter on their high megapixel cameras. The medium format backs don't have AA filters, but pros who use these backs shoot a lot more fabric and architecture where moire can be a problem than Canon and Nikon users.

The only hesitation I have about the modification is that MaxMax removes on one of the AA filters. I would like to get rid of the filter altogether, but apparently one of them is glued to the sensor.

I would be willing to pay Canon a premium for an 'L' (for landscape) camera with no AA filter to go along with their 'L' lenses.

ErikKaffehr · « **Reply #2 on:** April 22, 2009, 12:09:17 am »

Hi!

Check this thread:

http://luminous-landscape.com/forum/index....showtopic=30758

Erik

Quote from: markhout

Was curious about the effects of removing the AA filter to gain some extra perceived resolution and minimize capture sharpening. Came across this site: http://www.maxmax.com/nikon_d200hr.htm

Can anyone report on the practical pros and cons? I am shooting mainly landscapes, so I'm not overly worried about moire.

Thanks,

Mark

marcmccalmont · « **Reply #3 on:** April 22, 2009, 04:21:34 am »

I had the original 5D MaxMax and with 13mpix removing the AA filter did increase the sharpness (really only removing one AA filter, Canon uses a horizontal and a vertical) It is expensive $450. Now with a 5DII (21mpix) the sharpness is there and I feel no need to have it modified (probably can't because of the dust reduction) so if it were my money I'd put the $450 towards a 5DII
Marc

Plekto · « **Reply #4 on:** April 22, 2009, 05:03:04 pm »

IMO, it's always better to have no AA filter since you have to process on your computer anyways - and applying an AA filter via software(or trying several different ones) is always better in the end than the quick and dirty methods cameras use. Without a doubt better, but you have to do it yourself. No quick and simple here(which is what almost everyone wants other than artists and pros so it seems)

The AA filter being removed can also get rid of some minor optical issues as well, since there's nothing in between the sensor and the lens.

Lastly, if you bracket and blend for HDR, AA is your biggest enemy. You want the blending software to do the AA for you and not have to fight the AA that's native in the image. Side-by-side examples of HDR with AA and no AA of the same scene are obvious which works better.

Kitty · « **Reply #5 on:** April 22, 2009, 11:24:52 pm »

When removing AA filter, they replace any filter infront of the ccd?
If not how to clean up ccd when there is dust?

kitty

marcmccalmont · « **Reply #6 on:** April 23, 2009, 02:00:54 am »

Quote from: Plekto

IMO, it's always better to have no AA filter since you have to process on your computer anyways - and applying an AA filter via software(or trying several different ones) is always better in the end than the quick and dirty methods cameras use. Without a doubt better, but you have to do it yourself. No quick and simple here(which is what almost everyone wants other than artists and pros so it seems)

The AA filter being removed can also get rid of some minor optical issues as well, since there's nothing in between the sensor and the lens.

Lastly, if you bracket and blend for HDR, AA is your biggest enemy. You want the blending software to do the AA for you and not have to fight the AA that's native in the image. Side-by-side examples of HDR with AA and no AA of the same scene are obvious which works better.

All sensors have a UV/IR filter over them anyways (even MFDB) so there is always some glass over the sensor. The photosites are never "uncovered". Canon uses a horizontal and a vertical AA filter the rear one cannot be replaced only the front one.
Marc

Ben Rubinstein · « **Reply #7 on:** April 23, 2009, 04:59:34 am »

Quote from: Plekto

IMO, it's always better to have no AA filter since you have to process on your computer anyways - and applying an AA filter via software(or trying several different ones) is always better in the end than the quick and dirty methods cameras use.

What utter nonsense. Software filters don't come close, can't fix half of the problems anyway, etc.

markhout · « **Reply #8 on:** April 23, 2009, 07:20:55 am »

Quote from: marcmccalmont

I had the original 5D MaxMax and with 13mpix removing the AA filter did increase the sharpness (really only removing one AA filter, Canon uses a horizontal and a vertical) It is expensive $450. Now with a 5DII (21mpix) the sharpness is there and I feel no need to have it modified (probably can't because of the dust reduction) so if it were my money I'd put the $450 towards a 5DII
Marc

So Marc - just to be clear on this - if the only reason for upgrading to a 5DII would be the perception of sharpness/resolution, why not just pay $450 ? I have a D300, virtually worthless as a used sale, with which I'm perfectly happy, so an upgrade for $450 seems to make sense economically.

Plekto - I assume you mean that no AA would work better for HDR, right?

Thanks all!

Mark

marcmccalmont · « **Reply #9 on:** April 23, 2009, 11:38:11 am »

Quote from: markhout

So Marc - just to be clear on this - if the only reason for upgrading to a 5DII would be the perception of sharpness/resolution, why not just pay $450 ? I have a D300, virtually worthless as a used sale, with which I'm perfectly happy, so an upgrade for $450 seems to make sense economically.

Plekto - I assume you mean that no AA would work better for HDR, right?

Thanks all!

Mark

No what I meant to say is: I would not invest $450 on a 5D to make it sharper I would rather put that money towards a 5DII (or Sony A900)
The 5D II has greater resolution, more DR, greater bit depth, liveview, microfocus, better menu system, dust reduction. All things considered a better camera. I sold my 5D maxmax for $1300 and bought my 5DII for $2700 so the upgrade cost was $1400. If you want more apparent sharpness and more micro-detail spend the $$$ on an new body not removing your AA filter. Since you have a D300 I would wait for the FF Nikon equivalent to a 5DII I'm sure in a year they will have a $3000 FF 24mpix body to compete with the 5DII and A900.
Marc

markhout · « **Reply #10 on:** April 23, 2009, 03:04:55 pm »

Quote from: marcmccalmont

Since you have a D300 I would wait for the FF Nikon equivalent to a 5DII I'm sure in a year they will have a $3000 FF 24mpix body to compete with the 5DII and A900.

Point taken. But the used value of my D300 is just a few hundred $$$, so assuming that the new Nikon FF will sell for $3000, it would still cost me $2700 or more. Granted that there will be better DR etc, but then there is the wait - I believe you're correct in saying that it will take at least a year.

So all in all $450 for an interim update of my D300 (presumably they would clean the camera as well) is perhaps not a bad idea.

Thanks for the input.

Mark

Plekto · « **Reply #11 on:** April 23, 2009, 06:06:44 pm »

Quote from: pom

What utter nonsense. Software filters don't come close, can't fix half of the problems anyway, etc.

Most people here, though, find that their shots without an AA filter are fine is-is and don't do more than the very minimum of AA processing anyways. Even then, only if it's noticeable. But most things that you examine that carefully could easily be bracketed a tiny bit and blended(landscapes and static indoor shots and so on). At least there's a choice here, depending upon what you want.

As for bracket-and-blend HDR, AA filters are a problem. Because the HDR software (like the one frequently discussed here on this site) does its own anti-AA and cleaning up of the image. It won't hurt things terribly much, but it will be double-processing and take off some of the detail.

Lastly, the real gem in their lineup is the Fuji because it does in-camera HDR out of the box(two sets of sensors, then blends them into one image). If you remove the AA filter, it helps the camera a lot. But it also means the camera is only good in "High Quality"(double sensor) raw mode - the high ISO and high resolution modes, which don't use the double sensor mode, will have tons of artifacts.

The results, though, of the S5 Pro with the AA filter removed look almost identical to the Sigma SD14. Film-like smooth and clean in that one mode. Just without the iso problems of the Foveon/Sigma sensor. The image is also a tiny bit larger than the SD14 as well. The only downside is the thing is still very expensive used. Add in the cost of the conversion and you're approaching half the price of a new A900, which as others have stated, will crush any of the older generation cameras quite easily(and which will of course be crushed in a few years itself and so on...)

edit:
I guess that's my only real gripe with the company. They do great work but their offerings and models that they can modify is almost an entire generation old now. 8-12MP just isn't cutting it any more.

Toby1014 · « **Reply #12 on:** April 24, 2009, 04:00:21 pm »

Quote from: Plekto

edit:
I guess that's my only real gripe with the company. They do great work but their offerings and models that they can modify is almost an entire generation old now. 8-12MP just isn't cutting it any more.

5DMKII:

http://www.maxmax.com/canon_5d_mkii.htm

Anti-dust system is disabled on their conversions

Phil Indeblanc · « **Reply #13 on:** December 25, 2009, 09:03:30 pm »

Quote from: Plekto

IMO, it's always better to have no AA filter since you have to process on your computer anyways - and applying an AA filter via software(or trying several different ones) is always better in the end than the quick and dirty methods cameras use. Without a doubt better, but you have to do it yourself. No quick and simple here(which is what almost everyone wants other than artists and pros so it seems)

The AA filter being removed can also get rid of some minor optical issues as well, since there's nothing in between the sensor and the lens.

REMOVE the AA filter!! At least give us a choice..... CANON!
Your selling a $3K to $8K dollar camera...its pro? than let me do my processing....We are not all wedding photographers here! I think 90% of people using this forum shoot RAW, as just about all professionals that dont do events.(its a guess).

I have the Kodak SLR Pro, and the images are very noticably sharper(vs. Canon), more 3D quality to head up against MFdb look (14mp). I always resort to wanting to use it, as I dont get the same sharpness 3d quality with my 1ds2, so I take it out and start to use the Kodak.... Then the magenta highlights color issues starts needing taming, and the ISO is hard to get to and resets ...not very flexible... it does go way down to "6" ISO, but within a relation to exp time. My workflow breaks up, and it just doesn't handle slips, or lighting towards the lens very well. Highlights are tint of magenta and a process(painfull sometimes) to remove depending on what colors subjects I shoot.

I really wish this thing was refined, as it is the closest to a MFDB quality. It even had a decent software and tether shooting....But I put both away and use the Phase One on important work that needs that 3D sharp pop.

If the Canon/Nikon/Sony would provide a body with option to remove both V&H filters, I think the bridge between MFDB and DSLR's would be even closer limited by all else besides sensors.
I hope Canon does it, as that is the format I have most glass in :-)

Jonathan Wienke · « **Reply #14 on:** December 25, 2009, 10:09:03 pm »

Quote from: Plekto

IMO, it's always better to have no AA filter since you have to process on your computer anyways - and applying an AA filter via software(or trying several different ones) is always better in the end than the quick and dirty methods cameras use.

Most definitely NOT true. To prevent aliasing, frequencies above Nyquist must be filtered out before sampling. Once the signal is sampled, the frequencies above Nyquist are aliased to lower frequencies below Nyquist, and there is no way to distinguish aliased signal from real signal. Color moire can be removed by blurring chrominance variations into oblivion, but that does not solve luminance aliasing--there is no software solution that can fix luminance moire. The only only way to eliminate luminance moire is to filter the signal before sampling--an AA filter.

Phil Indeblanc · « **Reply #15 on:** December 25, 2009, 10:24:59 pm »

Quote from: Jonathan Wienke

Most definitely NOT true. To prevent aliasing, frequencies above Nyquist must be filtered out before sampling. Once the signal is sampled, the frequencies above Nyquist are aliased to lower frequencies below Nyquist, and there is no way to distinguish aliased signal from real signal. Color moire can be removed by blurring chrominance variations into oblivion, but that does not solve luminance aliasing--there is no software solution that can fix luminance moire. The only only way to eliminate luminance moire is to filter the signal before sampling--an AA filter.

I am not sure what to make of this statement...."there is no way to distinguish a aa signal vs real signal"....Can you clarify that?

You, having said the above, how does Phase and other MF backs do it(handle lum alias) without a AA filter?

ckimmerle · « **Reply #16 on:** December 26, 2009, 12:25:40 am »

Quote from: Phil Indeblanc

I have the Kodak SLR Pro...

The you know how bad the moire issues can be with a camera sans AA filter. I used the 14N for three years and experienced regular moire issues that were all but unrepairable. Sure, the offending colors were a fairly easy fix, but the luminance part of the moire was, more often than not, a lost cause. Color artifacting was another issue that was a pain in the arse to work on.

I'll gladly sacrifice a small bit of sharpness for the benefits of an AA filter.

ErikKaffehr · « **Reply #17 on:** December 26, 2009, 02:38:21 am »

Hi,

It has been discussed a lot before. The scientists say that OLPF is absolutely needed on any system where the lens outresolves the sensor. Many photographers feel that AA-less sensors give better sharpness, more 3D-effect and so on.

Technically speaking, aliasing can be seen as false resolution. On a resolution target this is often seen as contrast inversion the black and white parts change place.

A very clear example can be seen here:
DPReview test of Sigma DP2

The image above is taken from a Sigma DP2 using the Foevon sensor. The Foeveon sensors don't have a CGA (color filter grid array) and don't have OLPFs. Because the Foevon doesn't have CGA it will not show colorful moirés.

Color moiré can be removed using local desaturation of color. It has been said that monochrome moiré cannot be removed in software. Moiré is just one example of the aliasing problems, but the most famous one. Normally it matters little, except for certain types of shooting where repetitive patterns are frequent. Architecture and fashion come to mind. As Moiré is a repetitive pattern it is very possible that at can be reduced by applying filtration in the frequency domain. Such techniques are well known, but probably not applied to moirés.

The Leica M9 is very prone to Moiré under lab conditions but few problems have been seen in normal shooting conditions.

The problem goes away once we increase resolution. If the sensor has higher resolution than the lens the problem goes away. Increasing the fill factor also acts in the same direction. There would be little need for OLPFs on a 3 micron sensor, like the ones used on compact digital cameras.

As a side note, per pixel noise goes up with decreasing pixel size and DR goes down. If the resolution is normalized this effect will be smaller. The DxO mark takes this into account by having options for "screen" (per pixel) and "print" (normalized).

Best regards
Erik

Quote from: Phil Indeblanc

I am not sure what to make of this statement...."there is no way to distinguish a aa signal vs real signal"....Can you clarify that?

You, having said the above, how does Phase and other MF backs do it(handle lum alias) without a AA filter?

georgl · « **Reply #18 on:** December 26, 2009, 03:00:22 am »

We're talking about compromises. Ideally we need a sensor with a high photosite-density beyond lens resolution and downsample the image later to lens resolution.
Bayer-filtered-sensors are also compromises, 2/3 of the color information is just interpolated! Ideally (or scientificallly spoken) they would need an AA-filter which "blurs" the information equally over (4) colored photosites which are later used to interpolate color information. And then we downsample the image again to about half the resolution (1/4 pixel count) to get useful information on the 100% level...

These are all compromises but it has shown that in most cases, non-AA-filtered systems are superior in IQ because you can selectively handle (=hide) moire just where it appears in post. That's the typical workflow in MF - one image is selected and then processed to "perfection" - AA-filters in reality are a compromise within themselves, they destroy information below Nyquist in the whole image (even where moire/alaising wouldn't appear) and on the other side, aren't able to avoid any alaising-artifacts. They are the much better solution if you need acceptable results fast - while in certain, rare situations unfiltered systems create unusable results without processing.

But I wouldn't spend much money on removing the existing AA-filter instead of getting the higher-res-model anyway...

Phil Indeblanc · « **Reply #19 on:** December 26, 2009, 03:41:25 am »

Quote from: ckimmerle

The you know how bad the moire issues can be with a camera sans AA filter. I'll gladly sacrifice a small bit of sharpness for the benefits of an AA filter.

You using it for 3 years sure shows you have more patients than I do :-) I just kept getting magneta most of the time with my spill lighting, etc. Post processing large # of files for color wasnt fun, so I switched to the Canon.

I honestly have not experinced any moire using the SLR/c, as I dont shoot fabric. As far as architectural, I have not noticed any moire either as I rarely tested it outside the studio(as its more intended use). I don't know if it is the lens calibration, but besides the Canon lens (100/180L/50), I also used leica glass with it 60/100 el R's.

I don't know about OLPF, or Nyquist, etc and other scientific measures that prove one way or the other.. I know from my experiences and side by side test from the cameras I have used, and I would take a NON AA filter camera ANY day for the work I do.....as I can surely see the difference and it is not something you can get back by post sharpening. I just don't think that glueing an AA filter on a pro camera perminently is the thing to do.

I do appreciate the science explained, as much as I grasp of it, but when you look at product shoots side by side, that are intended for professional ads/ print, there really is no reason to try and convince someone to accept a inferior result, specially when they have seen both and know in the back of the mind that this can be more tacky crisp polished, 3D sharper. I don't expect these thoughts to be agreed with with all shooters, but those doing closeups or studio work, it is one of the reasons to use a MFdb.

Author Topic: Remove AA filter? (Read 23909 times)