With or without AA filter

Discussion in 'Casual Photo Conversations' started by jphotog, Feb 5, 2012.

  1. Ponder that one of the big dslr manufacturers would release a 30+ Mpix small format (full frame) camera in two versions: one with and one without a high pass filter (anti-aliasing filter) in front of the sensor. Some people choose to remove the AA-filter. Leica sells small format cameras with no AA filter. Most medium format cameras come without AA-filter and so does the Sony NEX-7. The obvious advantage is the improved detail at the risk of moire-patterns (that might be possible to correct in post).
    Does anyone have practical experiece with removing AA-filter or cameras coming without it?
    Which version would you guys choose?
  2. Am I really reading this?
  3. Oliver, is that a philosophical question related to extentialism or are you trying to tell me that it is not of interest? I have not felt the need to study Sartre since I left school a long time ago, but I am really curious about other peoples take on to AA-filter or not.
  4. maxmax will remove the AA filter
    this page has better examples
  5. I took the IR Blocker from my D50 to allow IR photography. Replaced it with Quartz glass. Not sure where the AA bit fits in the sensor sandwich. Maybe I should try to shoot some fabric and see how it Moire's?
  6. I used a Kodak SLRn for several years, and have since had a Canon 1Ds, Nikon D2x, and now Canon 5D II. The SLRn did not have an AA filter, so while it was a 14 MP camera the resolution was more like a 16-18 MP camera, like the 1DsII.
    The problem is the moire effect. Kodak was driven out of the DSLR market by the huge internet rebellion over the moire effect. I do not know how the medium format digital backs have been left alone by the critics. I had enough moire problems to render 5-10% of my images unuseable. I highly doubt that post-processing could fix the problems that I had and even if it can it must reduce resolution more significantly than an AA filter.
    I still love the images that I took with the Kodak SLRn but the 5D II with it's AA filter has been a vast improvement, and with no moire effects. I'll take the AA filter please.
  7. From what I've read from MF digital users, moire is NOT correctable. If it hits, you area screwed. For a "general" outdoor photographer like me, the filter is probably a very good thing.
    Kent in SD
  8. I have a Kodak Pro14NX and have used a Hasselblad H3d. In both cases the camera does not have an AA filter, I use the Kodak in the studio and for doing landscapes and have not had an issue with moire in prints. You can see it on screen sometimes but I have yet to see it in a print.
    The H3d has yet to have moire rear its ugly head.
    As to the OP I would doubt very much that a camera company would muck up there assembly line by having two versions of one camera.
    Also it is my understanding that as the MP go up the chance of moire becomes greater given the same size sensor.
  9. Most MF digital backs either omit or make AA filters optional. I have a CFV-16, which has the same nominal resolution as a D3 but is nearly twice as sharp. This comes with some baggage. Scenes with fine, repetitive patterns, like fences, bricks or fabric, tend to exhibit Moire patterns which cannot be removed after the fact. This is never a problem with my DSLRs, nor is it a problem with photos that don't show the "hand of man."
    There are photos in my gallery (the grain elevators) taken with MF digital and a D2x which illustrate the problem.
    There are firms which will remove AA and IR filters from the sensor. I'm considering this conversion for older cameras I no longer use regularly or as backups. For about $500 you can have the IR filter replaced with one which blocks visible light. This allows you to use the viewfinder and TTL metering while shooting IR into the near-micron spectrum. Removing the AA filter, while keeping IR blocking, would be suitable for a dedicated nature or landscape camera. Excessive IR sensitivity in visible light shooting makes it hard to get neutral blacks, or shadows in incandescent light.
  10. Sorry Jonas, I did not mean to be very rude, it is just that the way you asked the question made the topic of this thread sound very similar to the one just below yours ("one of the big dslr manufacturers", 30+ mp full frame camera, optional anti aliasing filter etc.). If yours is a question about AA filters in general, that is different and I apologize.
  11. I don't take pictures of screen doors or fabric so I have never had a moire issue. I would prefer to have no AA filter. Also, there is software to remove or reduce moire in post production, just google it.
  12. Hasselblad Phocus software for the CFV back has Moire mitigation software. It produces artifacts that are nearly as objectionable as the Moire it perports to remove. Artifacts are unavoidable, so if there are alternatives to Phocus, you may find some better suited to certain subjects than others.
  13. "Also, there is software to remove or reduce moire in post production, just google it..."
    Perhaps that statement should be phrased as "there is SW available which purports to remove ...".
    I've tried a bunch of it, as well as DIY approaches, and, while it's easy to remove color moire, it's a time-consuming PITA to do a good job on removing luminosity Moire patterns. In fact, there was even a very detailed thread on the subject a year or two ago on photo.net, in the wedding forum: http://www.photo.net/digital-darkroom-forum/00W8gC
    It's much better not to let Moire happen in the first place. Fortunately, as Kent and Walt pointed out, natural scenes have a lot of random high spatial frequency elements (eg, leaves on distant trees, grass, etc.), and these just don't produce Moire patterns.
    Tom M
  14. For the moment, even though I am upgrading from my Leica M8 to another model, I have decided to keep the M8 for the time being as its lack of an AA filter and very lttle other intrusion between the scene and the sensor allows it to have an IR response that makes for interesting infrared images. Some have the AA filter removed from other cameras to allow this capability. Apart from other problems (not at all frequent) like Moire patterns, the lack of the AA filter apparently gives the M8 a higher resolution image than it would otherwise have had.
    If one can add an AA filter to an MF or other digital camera at a later date, maybe that would be an option.
  15. Jonas, it's a relief to see someone begin a rumor-related discussion with a specific premise that encourages a constructive discussion. If every rumor-ware thread began on such a constructive basis it'd be far more interesting since it helps clarify the pros and cons related to a rumored feature.
    For what it's worth, my own experience is limited to comparatively ancient technology, the Nikon D2H. The less aggressive anti-aliasing filter offers superficially satisfactory "sharpness" at the expense of ugly jaggies in diagonal lines and fine details such as hair. In actual practice that's been far more common than moire. But it's enough to persuade me to opt for a more effective AA filter next time and rely on software to attain the desired sharpness.
  16. I've contemplated this issue more than a few times, although I've neverf actually used a camera with no AA filter. I think there is a major division in the sorts of images we, as photographers, shoot -- man-made and natural. Man-made subjects often have regularly repeating patterns (e.g. grates, grills, screens, fabrics, and other patterns), and they are therefore subject to moire. Natural subjects generally have no such patterns, although that's not always true. Many animals create structures with regular patterning, e.g. a bee hive or a scallop shell. Other animals (and plants) may have regular patterning of their own. Consider a palm leaf or the elaborate pin feathering of many fowl. Even consider a simple feather. So I don't think it's a 100% safe assumption that 100% nature photographers would never experience a moire pattern without an AA filter.
    As to the software solutions for moire, I'm with Tom Mann on this one. I've been dealing with aliasing issues from the earliest days of sound digitization, even before CDs hit the markets, with the earliest waveform analyzers and digital oscilloscopes. I provided major technical input for the writing of a grant for the digitization of an important biological sounds archive before there was even any standard for digital recordings. These were the very sorts of issues we were exploring. I can assure everyone that the only possible post-processing solutions for moire would be kludgy and marginally effective at best.
    The OP has invoked an additional question, though, namely whether moire could be a problem on a 30 MP camera. I would say it would depend on the lens. Most lenses are not going to have the sort of resolving power that they can't be outresolved by the 30 MP sensor. This would especially be true of a 30 MP APS-C camera. As such, the lens "becomes" the AA filter. Even in this case, it would make no sense to remove the AA filter, as there would be no appreciable gains (over the lower resolution lens) anyway. The only time aliasing (moire patterns) would be at issue would be when the maximum spatial frequencies of the lens' projection would exceed roughly twice (in practice mayber 2.3x) the pixel pitch of the sensor.
    I predict that megapixel wars will progress to the point of being somewhat silly, such that our sensors will eventually outresolve even our best lenses anyway. At that point, we won't be talking about AA filters anyway, because the manufacturers will omit them. Meanwhile, I'll happily take my cameras with AA filters.
  17. Medium format lenses have greater resolution than any digital back, and will continue to maintain this margin for the foreseeable future. A 30 MP back has less than 40% higher resolution than a 16 MP back (square root of pixel count), or about 60 lp/mm. It would be a pathetic MF lens that had such a low capability. In order to make the back the limiting factor would require the lens have 4 times this resolution, but that's a separate issue. Any resolution equal to or greater than the spacial frequency of the sensor could lead to Moire.
    While many natural features have a fine spacial frequency, they do not normally extend over a significant part of the image. With cloth, it is usually sufficient to drape it with folds and other variations in order to disrupt the regularity leading to Moire. It is less problem with short lenses, which exaggerate perspective, which also disrupts this regularity.
    Another method which might forestall Moire would be using a diffusing or soft focus filter. A Softar lens has the property of diffusing highlights while maintaining detail. If anyone has tried this, speak up. I have the resources, and will pass on any results when I find the time to experiment.
  18. Although I haven't tried it, I doubt that a Softar or similar soft focus filters will completely eliminate Moire, but such filters will likely be better at it than cheap highly scattering, generic "diffusion" or "fog" filters.
    The reason is that one can think of the way such filters work as superimposing a set of images with fairly large angular changes over the original image. By "fairly large", I mean that when the deviated rays from these filters are passed through the lens in use, the offsets at the film plane are almost always much larger than a pixel or two. Because of this, the deviated rays will obviously not exhibit any Moire, but the undeviated ones will still show the problem.
    The question is then what is the fraction of the incident light is deviated by the filter. This depends its exact design. You always need to be able to recognize the original image, so you can either have (1) a filter where lots of the incident rays are deviated, but the maximum angular deviation for any particular ray is small (eg only a few pixels), or (2) you deviate a smaller fraction of incident rays, but make their angular deviation larger (eg, a significant fraction of the image) so you get more of a veiling fog. What you can't have is (3) most of the rays deviated by large angles, because in this case, you likely would have a hard time recognizing what you're looking at. ;-)
    To reduce Moire most effectively, you want most of the rays to have been deviated, even if only a few pixels. IMHO, this is most likely to happen, as Edward suggested, in a Softar type filter, not a fog-type filter.
    Tom M
  19. With or without the AA filter? I don't know - which one would let me drink more? ;-) I haven't used a filter free sensor. The side by side results look impressive, but so do side by side results with and without proper sharpening. Doesn't sharpening reverse the filter's negative impact?

Share This Page