Tuesday, October 11, 2016

At last! A real-world example of Induced Moiré


A lot of the world's high-end cameras have been boasting about the lack of optical low-pass filters in front of their sensors.  And while most hail this as a wonderful thing to help increase the detail in your images. it comes with a theoretical downside: your images might become more susceptible to a phenomenon called "induced Moiré".

Moiré interference patterns happen when two similar, regular arrays are not quite aligned properly.  In early television, Moiré patterns would show up a lot when cameras pointed at presenters wearing pinstripe suits or even finer-lined patterns.  The interference came from the interaction between the clothing and the 525 lines in the TV camera.  In color TVs the interference patterns had a lot of color burst associated with them as well.

Two screens of a similar square size (“frequency”) will develop an interference pattern when one is slightly misaligned with another (left).  Since the pixels in a sensor are arranged in an array similar to a screen, it’s possible that such a “induced Moire” pattern can develop when photographing another array of a proportional density (right).  Sometimes the effect is accompanied by color bursts.  (Right image courtesy Foveon.)  
Early digital cameras would have suffered the same fate had designers not tried to avoid the problem by installing what’s called an anti-aliasing or low-pass filter in front of the sensor.  Without going into the mathematics behind the design, it can safely be summarized by saying “the filter fuzzifies the image a little bit, eliminating the annoying interference patterns and making both consumer and professional photographers very happy”.  (Of course I’m oversimplifying this a lot.)

Example image courtesy Wikipedia.
Time passed.  Pixel densities increased.  A higher pixel density is the equivalent of a finer-pitched screen.  Finer-pitched screens don’t induce as much Moiré (well, it all depends on the subject being photographed, but in general it won’t produce as much) and so manufacturers have been experimenting with taking the anti-aliasing filter away, thus not intentionally fuzzifying the image and allowing all the detail of the subject to be properly captured.

Over the last few years as part of my explanation of this feature in my books, I tried very hard to actually create an image where such induced Moiré occurs, and, try as I might, I have never been able to create such an example.  (And so I always ended up using exaggerated examples from companies like Foveon which you would never actually see in the real world.)  

Then last month I received an email from a reader:

Dear Gary,
due to your great eBook about the Sony a7R2 I’m confident you know a lot of about the camera and how the camera works. [...]  To cut a long story short, I’m living in Beijing and made some pictures of the Forbidden City while I was on the Jingshan Park behind it. 

I used a Sony a7R2 and a Zeiss 16-35mm lens. I took some pictures of the Forbidden City and recogized some serious image quality issues in the area of the roof of some of the buildings.  I would like to ask you if you could tell me why the roof looks that bad and how I can avoid these kind of effect.
You are the only one I know who understands how this camera works and probably can explain to me what happend.  The Sony feedback service proposed me to clean the lens
Thank you very much in advance!  D. Winkelmann, from Germany

I had a look at his attachment, and was gobsmacked - the worst case of Moiré I had ever seen in a modern digital camera.  The red rectangles show off the problems pretty nicely (click on this or any other image to see it larger):


But how was he able to get such a great example when I had tried and failed to get examples like this for years?  The answer came later when he sent me the original, uncropped image:

Yup!  He had to pixel peep just to see his subject. :-)  Of course this is one advantage of high-megapixel cameras and high-resolution optics, but it instantly explained why my previous efforts to create an example had failed: I had misunderestimated (to quote a presidential term) the actual distance between pixel rows and columns on the sensor.  Just look at how close the rows and columns need to be in the image in order to produce this undesirable artifact!  And look at how closely you had to examine the image in order to see it!

So there's a very good chance you'll never see this problem in your images.  But what can you do if you’re unlucky enough to have it?  Well, the Sony guy's recommendation of "clean the lens" certainly won't help (although in theory making the lens dirty might fuzzify the image a little bit, emulating the effects of the anti-aliasing filter).  In the real world, Adobe Lightroom has a tool for reducing the rainbow artifact and you can see examples of how to use it here.  Getting rid of the actual interference pattern is more complex and you need a stronger tool such as Photoshop.

In the Pipeline

An earthquake in Japan which damaged one of Sony's sensor production facilities has made for a quiet summer.  I used the time to work on other projects - a new website for Xaphoon.com, and another project I'll talk about in a few months.  And now things are going to get busy again as Sony has announced 3 new cameras, all of which will have their own e-books: The A99 II (the new A-mount which varies slightly from the prediction I made more than a year ago), the A6500, and the RX-100 V.  (Send me an email at Gary at Friedman Archives dot com if you'd like to be on the notification list and have pre-release access.)

Of the three cameras, it is the recently upgraded RX-100 V that might have people scratching their heads.  Significantly more processing power and a new sensor with phase-detect autofocus built right in, shooting 24 images per second?  Sure, it's yet another engineering marvel, but what point-and-shoot owner needs that kind of power?

My guess is "very few"; however Sony needed to develop this improved sensor in order to make the upcoming-but-not-yet-announced RX-10 IV a powerhouse.  Recall that the existing RX-10 III is just an RX-100 with a really big 24-600mm lens slapped in front, and at such long focal lengths the contrast-detect autofocus makes it a weak choice for shooting sports.  But with this new PDAF sensor and powertrain, this humble little do-everything camera will be a serious contender for all but the most highly-paid sports photographers and videographers.  No longer will you have to spend $5-10K on high-end equipment to shoot close-ups of moving athletes and have publishable images!  Although I have no knowledge of Sony's future product plans, the RX-10 III was the camera that was in need of PDAF the most and it's clear to me that the new sensor and engine was designed with that camera in mind.  I truly believe that camera will be a game changer when it's announced.

The Fujifilm X-T2

Tony Phillips is typing his fingers off in order to get the X-T2 ebook finished.  Tony is my equal in terms of having the rare combination of writing skill, technical prowess, and the ability to explain complex things in an approachable way, and his books are so good that even Fuji has taken notice and have helped him get early access to cameras.  (Trust me when I tell you that not all companies do that. :-) )  Here is one of his latest videos showing the highlights of the X-T2:



If you're a Fujifilm fan, you can help Tony and me by helping to spread the word about his ebooks online.  Word-of-mouth from extremely happy customers is our only form of advertising!

A6300 in French!

Un autre Archives Friedman Premières - mon ebook sur le Sony Alpha 6300 a maintenant été traduit en français! S'il vous plaît aider à répandre le mot. Vous pouvez acheter votre propre copie ici.

Seminars – Arizona, Colorado Springs, Edmonton

The Friedman Archives Seminars schedule is shaping up for 2017.  Here's what we have so far:

  • Tuscon, Arizona - February 14-15.  (Photo club members only.  If you want to attend, you have to join their club. :-) )
  • Colorado Springs, Colorado - February 18 - 19.  
  • Edmonton, Alberta (Canada) - sometime in May.
We will do two more seminars in the 2nd half of the year to any photo club that requests it.  If you're a member of a photo club, please ask me to send you an introductory packet.  And if you're interested in attending any of the above, please fire off an email.

Can't make any of the above?  You can intuitively understand and untangle the complexities of digital photography in a fun and enjoyable way, no matter where you live in the world, via the streaming version of the seminars.  These are, truly, the shortest path to "Wow!".

f2 Cameracraft 

You've heard me talking about f2 Cameracraft for a long time.  The current issue might not make it past the censors in Dubai, but you can get yours by subscribing now.

f2 Cameracraft is designed in the tradition of Minolta Mirror and National Geographic magazine - perfectly executed printed editions that are so good you can't bear to throw them away.  (Well, the National Geographic comparison is a bit unfair - our print quality is WAY better than theirs. :-) )  Not sure if these bi-monthly collector's magazines are for you?  Have a free look at the previous edition and decide for yourself!

Parting Shot

The shots below were taken from a recent trip to Ferndale, California.

The shortest lighthouse I ever saw.

The same lighthouse, taken on a night so clear you could see the milky way overhead.  Normally you can't mix such a bright object and such a dark background, so I too two shots and merged them.  If you look really carefully you'll see a shadow of my A6300 on a tripod and me doing my Nixon impersonation.

Until next time,
Yours Truly, Gary Friedman

24 comments:

  1. The only time I've been able to induce moire was on my a700, I was photographing United Airlines' blue color scheme and the tail had a slight rainbow tinge. Their tail at the time was basically a very thinly banded gradient. It was subtle, but noticeable and not pixel peeping either. This is on a camera with an AA filter, too! It cropped up a few times, but not all times.

    http://www.airliners.net/photo/United-Airlines/Boeing-757-222/1721474

    ReplyDelete
    Replies
    1. That's interesting because the A700 HAD the fuzzifying filter installed, which was designed to minimize this. Oh, well. :-)

      Delete
    2. If you could get up close and personal with those United tails you would see that they were all composed of strips that had the subtlest gaps for the blue to show behind. The gaps formed this grid pattern that you could only see if you were standing up close to the jet. It only seemed to happen at certain angles of incidence too, which is also true of most moires as well. I have plenty of blue-livery Uniteds that don't show any moire. Just a lucky coincidence!

      Delete
  2. RE: A99ii
    WOW, you called it. I am really glad the translucent mirror stayed intact. It's one of the reasons I like the A99. And, I have to amend my statement from back then "at my age I may never buy or need (famous last words) another camera". After seeing the specs on the A99ii I'm saving my pennies. :-) The A99ii will definitely be the last one. :-)

    ReplyDelete
  3. I do have an example from one of mine on a sidewalk oddly enough - if you want I can try to dig it out and send it.

    ReplyDelete
  4. Hi, Arthur. Thanks for the offer, and yes, I'm curious to see it! GF

    ReplyDelete
  5. I'm quite surprised at your difficulty in finding examples of colour moire. I have an A77ii and with one of Sony's plastic 50mm primes I have problems with this frequently. I like to take photos of architecture. Examples are metal fences with vertical bars or the corrugated metal that is often used at the top of city tower blocks. And tiled roofs like your example. It can be quite visible even at full-screen size rather than 100%.

    Colin

    ReplyDelete
  6. Hi. How can I be notified when the XT-2 book is available?
    Thanks!!!

    ReplyDelete
    Replies
    1. Send me an email with the request and I'll add you to the notification list. GF

      Delete
  7. Great to see that Sony have announced the A99ii - I thought A mount was dead, but I'm definitely saving up to get this for Christmas!!

    ReplyDelete
  8. Have you written anything regarding the Sony e-mount 16 mm and 20 mm wide angle lenses, with side remarks about other e-mount wide angle offerings. For the a6300 size sensor.

    ReplyDelete
    Replies
    1. I haven't, only because there are plenty of other websites that obsess about these sorts of things, so there's no need of adding to the noise. In my opinion great optics are meaningless without great light. (See the previous blog post.)

      Delete
  9. Hi Gary, Are you going to look at the new Olympus EM-1 ii?

    ReplyDelete
    Replies
    1. Tony Phillips might. If course they'll have to formally announce it first. Send an email to register your interest! GF

      Delete
  10. Hi Gary! Does the new A6500 have an AA filter? Tony Northrup says yes, but I haven't been able to verify from anyone else. Thanks

    ReplyDelete
    Replies
    1. It wasn't officially stated but I agree with Tony's guess - so far Sony has only removed the filter on full-frame sensors.

      Delete
  11. Thanks Gary. Your technical insight adds an unusual and valuable dimension to your writings!

    ReplyDelete
  12. Sometimes using a different demosaicing algorithm diminish or even prevent moiré.
    Clear examples with the RAW files at http://forum.luminous-landscape.com/index.php?topic=104708.msg862405#msg862405

    When resizing images the Bicubic or other poor implemented algorithm may also creates moiré.
    Example https://dl.dropboxusercontent.com/u/4686872/FastStone_Photoshop.gif

    A BIG thread disusing this http://forum.luminous-landscape.com/index.php?topic=91754.0
    The *last/best* script from the thread https://dl.dropboxusercontent.com/u/4686872/Resise_V122.bat
    Put the script into the Desktop and drag&drop the files over the script icon.
    The script uses ImageMagick version #6, the latest Win version 6 ftp://ftp.imagemagick.org/pub/ImageMagick/binaries/ImageMagick-6.9.6-2-Q16-HDRI-x64-dll.exe
    The script also deals with the gamma problem http://www.4p8.com/eric.brasseur/gamma.html

    Examples of moiré produced by Photoshop Bicubic vs Resize Magic
    http://www.fsoft.it/Imaging/en/Esempi.htm
    Resize Magic is interesting, better than Photoshop Bicubic, but also
    may creates some ringing artifacts...

    ReplyDelete
    Replies
    1. Thank you, Pictus, for this wonderful collection of resources!! GF

      Delete
  13. "It is the recently upgraded RX-100 V that might have people scratching their heads. Significantly more processing power and a new sensor with phase-detect autofocus built right in, shooting 24 images per second? [W]hat point-and-shoot owner needs that kind of power?"

    Moi, I do. I will be buying the RX100 V along with Sony's new underwater housing as a replacement for my Nikon 1 AW 1, which is starting to show its age and its limitations. The Phase Detection AF of the Mk V is a necessity.

    ReplyDelete
  14. How can I discuss optical low-pass filters, a theoretical downside, Moire interface patterns, and a higher pixel density equivalent of a finer-pitched screen. I cannot, but most of you reading this intelligently written blog can, and I am not jealous. I am glad to be counted among you and learning more with each publication, never hoping to reach the top, but certainly enjoying the journey.
    I thank you Gary for sharing with all of us on the learning scales.
    Sheila

    ReplyDelete
  15. Dear Mr. Friedman,
    I thoroughly enjoy your newsletter, and your most recent one touched on a topic that has confused me for a few years. You discuss the “anti-aliasing or low-pass filter in front of the sensor.” I have been into infrared photography for a few years, and am on my second converted camera. I am somewhat of a local “expert,” and have developed and presented infrared topics at local camera clubs.

    I know that when a camera is converted to infrared, they remove the “low pass” filter, which prevents infrared light from hitting the sensor and install an infrared filter that only passes some selected bandwidth of light above the visible spectrum. I have always assumed this included removing anti-aliasing function.

    I also know it is possible to remove the low-pass filter and not replace it with anything. If you do this, you can use the camera for color or infrared by attaching a lens filter that either passes low light or passes infrared.

    Now that cameras are being produced without the anti-aliasing function, I have not seen any mention of the low-pass function. Do you know whether there is still some low-pass filter or whether the infrared spectrum is allowed to hit the sensor in these cameras that have removed the anti-aliasing feature?

    Harry O’Connor
    oconnorhj@yahoo.com

    ReplyDelete
    Replies
    1. Hi, Harry! Before I answer let me explain to readers that "low-pass filter" is an ambiguous term that can mean either "anti-aliasing filter" as described above, or "infrared blocking filter", meaning all wavelengths lower than the infrared wavelengths will pass through to the sensor. You ask "If you remove the anti-aliasing filter, is the IR filter gone too?" The answer is no -- there is still a filter in front of the sensor that blocks IR wavelengths but lets the visible light through.

      Delete
    2. I figured that had to be the answer, but I've never heard anyone state it before.

      Reminds me of the old joke about the government standards office that had two departments: "The Department of the Same Name for Different Things," which was right next door to "The Department of Different Names for the Same Thing."
      Thanks.

      Delete