An Example of Real Aliasing

By Carl Weese

I bet the resolution of the D800, E or not, would dramatically change this situation. Here's a screen grab of the ACR window in 100% view looking at a file including a corrugated screen in glancing afternoon sunlight. I think that the distance/focal-length was such that the corrugation in flat light could have been at the limit of resolution, but adding a shadow from raking light for each "stripe" dropped it below pixel size so [in Carl's original —Ed.] we've got true sensor/subject pixel-level moiré interference:

(click to open)

The patterns on screen change at different magnifications, but they never go away. Based on the 100% view, I don't see how it can go away in a print—the basic structure of the subject has not been imaged, it's replaced by a bunch of false data—but I'll do a test.

The other side of this, however, is that I can easily see how more-than-doubling the sensor resolution could blow right by the problem, though of course the G3 does in fact have an AA filter, however weak.

It does make me wonder what else could set things off—not set off the annoying screen-only stuff, but actual pixel/detail interference, even with 36MP. Rows of windows of buildings on the Manhattan skyline from Hoboken? Cable bridge at just the right/wrong distance? Somebody's Harris tweed jacket in a street scene? Building in background with contrasty brick and mortar pattern?

Carl

Original contents copyright 2012 by Michael C. Johnston and/or the bylined author. All Rights Reserved. Links in this post may be to our affiliates; sales through affiliate links may benefit this site.

(To see all the comments, click on the "Comments" link below.)
Featured Comments from:

No featured comments yet—please check back soon!

Comments

[Rob Young]

Photographing in Australia I sometimes have issues with vertical corrugated iron used in farm sheds.

[Howard Sandler]

Aliasing moire is not all that uncommon in the studio. The Nikon D70, my first DSLR, had a weak antialias filter and was particularly prone to it when using a really sharp lens like the 85mm f1.8 at around f5.6, which is a common working aperture with studio flashes. Oxford weave shirts and tweed were particularly troublesome. I usually did not have time to zoom in to 100% and chimp to make sure it was not there, so I did a lot of colour cloning in post to reduce the look of the moire when necessary. The D80 was less prone. Now I use the D90 and see it occasionally. I hypothesize that with the higher resolution of the D800E it will tend to be seen at greater distances, as in group portraits of 4 people or so.

[Larry Gebhardt]

And if you convert this to black and white the moire almost disappears.

That's how the moire reduction filters seem to work. They just remove all the color and replace it with a single color. It's not a bad option for when something ugly like this pops up, but in this case the screen might be the subject, so it may not yield a satisfactory result.

[carlweese]

That's it exactly. What camera, what resolution (MP count) and what distance and lens length? The problem clearly seems to be having regular detail land right on, or just past, the limit of the pixel array. But a host of factors determine where that will happen.

[John Driggers]

I have had this issue in LR with the previews, but when rendered on export, the pattern is gone. What happens when you export the file?

[Harold Merklinger]

Maybe I don't understand your form works; now it looks like my first comment didn't go at all! It went something like this:

http://users.eastlink.ca/~hmmerk/SubtleMoire2.jpg
Here's a real example of moire, pixel-for-pixel, from a D800E. Note the subtle colour patterns in the roof of the blue building and in the roof just below it on the right. One pattern survives the conversion to B&W while the other does not. I also did a different conversion to B&W where neither pattern appeared in the B&W version. Nikon D800E, 24-70/2.8G, at 28 mm, f/9. The same scene taken at 24, 35 and 50 mm showed no moire. Leica M9 images of the same scene with the same focal lengths and aperture also showed no moire. I have not been able to produce moire on demand with the D800E, but it does appear - as here - when I would not normally expect it. Worst camera for moire in my experience is the Kodak DCS Pro SLR 1/c (14 megapixels, full frame no anti-aliasing filter).

[ctein]

Dear Carl,

The D800E might avoid this particular problem, but 30+ MP non-anti-aliased sensors can still produce it, especially when you don't want it (it's a Murphy thing).

When I had the loan of Jeff's Phase One P30+ back, I got some remarkable examples of primarily-chroma aliasing photographing the Palace of Fine Art. This occurred with structures that were 10's of pixels in size. Whodathunk. Somehow it just hit things wrong.

A more common situation is when you have a fine very dark sharp-edge feature against a brighter background. Trees and bushes can be great for this. Depending on just how the edge cuts across the Bayer array, you'll get all sorts of chroma variations along that edge. This doesn't require a small object, just a sharp transition.

Being B&W-oriented would be a distinct advantage in such a situation.

A mediocre lens can also help a lot! [smile... but true]

pax / Ctein

[Mark Kirkpatrick]

Hi Carl,

I empathize with your pain, and I share it all because of you. I was driving through Memphis this summer. Passing an abandoned drive-in, and having recently bought a couple of your wonderful prints, I was inspired to stop and take a bunch of frames. Oh whoa is me: they are a singular hell! It seems that your life's mission is a perfect digital moire storm.

Bon courage,
Mark

[Mark Kirkpatrick]

Another place where these artifacts live: a thin white line (on the order of 1 pixel in width) is rendered in changing hues along its length. This shows up in architectural scenes. Increasing the MP count only changes the scale where this artifact lives, but can't make it go away. One escape may be with sensors that don't use periodic color arrays (as Bayer arrays do).

[David Bennett]

I see moiré interference in twill shirts with the unaided eye. Anyone else see moiré interference other than as produced by a camera sensor?

[Mike Shimwell]

As long as the lens can resolve at a sufficient level you will sometimes get aliasing with non-AA's sensors. Sometimes it's shows up as colour moire, sometimes as 'funny' edges. I shoot with an M9 and that does it, handheld at quite normal apertures with various of my Zeiss ZM lenses. A D800e must perform in the same way, albeit with a higher initial frequency staring the aliasing. It's just physics.

I've also been playing with a pansonic GF2 and that and the GX1 do the same. My old Canon dslrs also aliased sometimes, but less so.

It's annoying when it shows up in a print as there's not really much you can do about it, without becoming the 'digital painter' and making it up.

Mike

[Ed Grossman]

Pardon the pun, but this has all been moire on the topic than I ever cared to learn.

Mike, I'm glad to see that you & the D800E are getting along so well. Make sure you keep the box & manuals from the one you order. I hear it helps the resale value *wink*

[Doroga]

Different raw interpolation algorithms can affect moire production.

So some other software could do better.

[carlweese]

B&W conversion in ACR can make the problem less obvious, but at least with the corrugated metal it doesn't actually fix it. Harold, in your example, the roof seems to have uneven discoloration in the b&w version, where the color moire used to be, but I don't know which is the chicken and which is the egg.

I think this is the sort of thing that needs such a specific set of factors that to most people it will seem random when it happens. But if you work with fabrics all the time, as Howard mentions, a consistent set of factors will show up. For me, the corrugated metal is a very common material in the drive-in theaters I was shooting this summer and fall so I had to be constantly on guard against this. Ctein, I found that intentionally "killing" the lens quality by stopping down past the diffraction limit could eliminate the moire while still leaving adequate resolution for most purposes.

[carlweese]

John, I don't use Lightroom. When opened from ACR into Photoshop, the 100% view remains exactly the same. I've tried using a variety of different sharpening routines, both capture and output, but most had little effect. Oddly, applying what I would normally consider way too much output sharpening seemed to bury the problem. It's possible that doing this locally could be a useful workaround.

[Andrew Molitor]

At some point the lens diffraction probably provides an adequate low-pass filter that the AA filter is unnecessary, but obviously only at apertures sufficiently small for whatever the underlying pixel size is. I have always been deeply suspicious of cameras that ditch the AA filter.

The world is different in 2 dimensions, but in the audio world a correctly implemented and good quality anti-aliasing filter degrades the signal essentially not at all, while preventing aliasing problems. Building a 2D brick-wall low pass is, I suppose, harder.

[Joe]

The absence of an anti-aliasing filter can lead to another artifact.

I haven't shot with the D800E, but I had the anti-aliasing filter removed on my D700 by a company in New Jersey that replaces the filters with optical glass.

I've since found that pinpoint white lights against black will show up as colored lights. I've seen this when shooting, for example, a distant bridge at night.

I imagine it's caused by the fact that each pinpoint is hitting an individual photosite, which is of course limited to recording a single color. An anti-aliasing filter would spread the tiny pinpoint to hit more than a single photosite.

[Mark Layne]

How about a bracketing rotating sensor - just like film halftones in the old days - if anybody still remembers platemaking from film

[Josh Marshall]

I predominantly see moire in suits as a wedding photographer, and my fix is to desaturate the area and paint back in the original colour with an adjustment brush. Can be done in seconds with Lightroom.

My most amusing experience with it however was helping another photographer with printing a family portrait. The original file showed no moire, but downsized and printed it did. It took some work to convince him that we had to blur the problem area significantly *before* downsizing... because of course it was being introduced by the resizing algorithm.

[Earl Dunbar]

Don't you just love film?

[Earl, Carl shoots lots of film...often 240 square inches at a time. [g] --Ed.

[carlweese]

As a follow-up, I printed this shot sized down to about 11x14 and was surprised to see the print had less moire than I'd expected. Not at all an acceptable print, but *some* of the nastiness was absorbed by the interpolation and dither. Makes me think that it might be possible, somehow, to retouch away enough of the effect to wind up with an acceptable print.

On film: because of the huge number of subjects I was shooting on the drive-in project, I had to ration the 8x10 and 7x17-inch film, shooting only digital captures unless *both* the theater and its setting were very distinctive. But several theaters got done with film primarily because the exact framing I wanted was clearly going to risk these moire problems.