Reading mode: Light Dark

Aliasing (moire) vs. Anti-aliasing (softness)

Go to last post
TurboTed said:
I'll take your word on that. It seems to me that hair is just
plain difficult regardless of your method. With no AA filter, I
would expect to see all kinds of colored artifacts in hair. Not a
nice perfect moire pattern but it would still have issues.
Click to expand...
Ted,

Agreed. The Fuji S2 has outstanding resolution and is well known for moire in the hair. The Nikon D2H is noted for its acuity and there doesn't seem to be a serious moire problem. As I was telling Patrick, each camera has its sweet spot. I just cannot accept the suggestion that the Mk2 is soft because it is better.

Regards,

--
Jim
 
Jim,

Reading your first post, I think you may be misunderstanding a point I made. The designer will trade off moire with softness, meaning he will adjust the strength of the AA filter to get more of one or the other. I was not saying that he will decide whether or not to have an AA filter. Since moire (or aliasing) is an effect that is far more destructive, I believe that the prudent and conservative decision is to lean towards softness, which can be compensated once digitized, either in-camera or in the PC. Ideally, the engineer should find the perfect bandwidth of a steep (and expensive) AA filter so that it is visibly sharp and moire effects are rare (the "sweet spot" you mention). Practically, one may find that cost, lens choice, subjects, and personal taste affect that decision, so the neat solution is to give the user lots of flexibility to play with parameters in-camera or in software to customize to the desired taste and subject. My suspicion is that Canon follows this type of thinking.

You made an interesting point that photos with motion blur and softness are difficult to sharpen. I find this surprising because motion blur is a spatial low frequency effect (i.e. over tens or hundreds of pixels), while antialias blurring is a spatial high frequency effect (over just a few pixels). Unsharp mask has the ability to pick which frequency you are enhancing with the radius parameter. A very close analogy is to say that my musical recording has intentionally strong bass but the treble is weak. With an equalizer, you can enhance the high frequencies while leaving the bass intact.

I am, of course, speaking from a theoretical, rather than practical viewpoint, because very few of my photos have intentionally introduced motion blur. If you say so from experience, I would readily concede that I am probably missing something. I just like to try to understand why things don't work as they seem they should.

Pat
Jim T H said:
TurboTed said:
I'll take your word on that. It seems to me that hair is just
plain difficult regardless of your method. With no AA filter, I
would expect to see all kinds of colored artifacts in hair. Not a
nice perfect moire pattern but it would still have issues.
Click to expand...
Ted,

Agreed. The Fuji S2 has outstanding resolution and is well known
for moire in the hair. The Nikon D2H is noted for its acuity and
there doesn't seem to be a serious moire problem. As I was telling
Patrick, each camera has its sweet spot. I just cannot accept the
suggestion that the Mk2 is soft because it is better.

Regards,

--
Jim
Click to expand...
 
Jason Hutchinson said:
Jason
Patrick Pai said:
Just read through a few posts regarding the 1D (mk I) when it was
first released in the fall of 11/2001. Interesting to note that
some people were pleased, but quite a few posts from people
concerned and upset about visible moire in early sample photos
posted on the web The tone is almost identical to what we are
seeing now, but on an issue that is the other side of the coin.

http://forums.dpreview.com/forums/read.asp?forum=1019&message=1801916

Would like to comment on this from an engineer's perspective. I've
some experience in signal processing and CMOS IC design. The terms
"aliasing" and "anti-alias filter" are commenly used in
discrete-time signal processing. When a signal is sampled (either
in time or in space), there are some unpleasant effects which can
occur. If you are interested, start searching on the web for the
details.

Anyway, to my point. Aliasing is when you have frequencies
higher than one-half of the sampling frequency. i.e.
photographing an object which images to, say, 101 lines per inch on
the sensor, while the sensor has a resolution of 200 lpi. (the
magical 100 lpi threshold is known as the Nyquist frequency). The
recovered image will have a visible "beat" (or moire) at 99 lpi.
You can't tell this alias from a real signal at 99 lpi. In fact,
if there was an image component present at 99 lpi as well as 101,
you couldn't figure out after sampling how much the original had
and how much is the alias. Once aliasing occurs, information from
the original signal is lost, and you can't recover it, period. You
can hide it to some extent, but you have lost some fidelity due to
aliasing.

To minimize aliasing, it's best to filter out the frequencies above
Nyquist before sampling at the image sensor. This is achieved with
a lowpass anti-alias filter with a cutoff frequency lower than
Nyquist. What's the tradeoff here? Sharpness (i.e. some of the
high frequencies above the lowpass cutoff, but just below Nyquist).
Note however, that this can be compensated for in the digital
domain with an equalizer (unsharp mask in PS). Note that
oversharpening results in ringing (halos). Why is a lowpass
anti-alias filter preferable to moire? Because there is much less
loss of fidelity than in aliasing.

Most of this has already been said from a photographer's
perspective. I realize that if you haven't been exposed to signal
processing theory this won't be as much fun to you either. Here's
the bottom line: when designing a digital camera, a designer will
have to trade off moire and raw-image sharpness. You can get more
raw sharpness, but at the cost of moire. And since the raw-image
softness can be compensated for when post-processing or during
conversion to jpg with much less loss of fidelity, it obviously
should be biased in this direction. If you want sharp photos out
of camera, set the parameters to sharpen up your jpg files
in-camera.

Pat
Click to expand...
Click to expand...
Although it is impossible to design a true step function filter (high or low pass) it is possible to have designs that have very steep rsponse but contain ripple in the pass and stop band. What is the effect of this ripple on image quality as opposed to the steepness of the transition from pass to stop band. What do the filter responses look like for real world A filters?
 
When I asked the Canon Rep about the Moire' problem I had with my 1D, he explained that there was greater than a 90% chance it would not show up on the print. Mostly it was only on the monitor and is only evident at some resolutions. He said if you view the image at 100 %, Moire' would not show up even on the monitor. He demonstrated this on his laptop at different resolutions.

One of the reasons I sold my 1D was the Moire' problem with brick buildings. I assumed since it was evident on my monitor, it would be evident on the final print. I didn't even bother to print the images.

To those of you who think you have a problem with Moire', I recommend you try the above: view at 100% and make a print to see if there really is any Moire'.

Ray Amos -
1v (waiting for my 1D2) and EF lenses from 17-500mm.
 
Razor sharp but with moire. Right?

Isn't that why there is some issue with the 1Ds and moire? The AA filter is "weaker"?
Jim T H said:
TurboTed said:
I'll take your word on that. It seems to me that hair is just
plain difficult regardless of your method. With no AA filter, I
would expect to see all kinds of colored artifacts in hair. Not a
nice perfect moire pattern but it would still have issues.
Click to expand...
Ted,

Agreed. The Fuji S2 has outstanding resolution and is well known
for moire in the hair. The Nikon D2H is noted for its acuity and
there doesn't seem to be a serious moire problem. As I was telling
Patrick, each camera has its sweet spot. I just cannot accept the
suggestion that the Mk2 is soft because it is better.

Regards,

--
Jim
Click to expand...
 
Patrick,

Nicely explained. Nyquist is always a bit mysterious
to folks, including some physicists like me who should
know and understand better. In particular
when you note things in real life such as the fact
that the best I have ever seen in edge fall off
in a digital image is at best 2-3 pixels 10-90% fall,
I knew the AA filter was the culprit but
never thought thru the Nyquist frequency argument.
Never thought about the Nyquist
frequency per se as the explanation of
the AA filter put in the body. But if you have an 8 um
pixel, a diffraction limited F8 lens in the midvisible with resultant
8 um resolution and it takes 2-3x that distance for the
instensity of a knife edge to drop, the answer pops
out at you. Not only must you be uncertain of locations
to at least + - 1 pixel, but the AA filter was probably designed
to limit you to less than Nyquist, thus better than 16 ums. Viola, if you
get 2-3 pixel transitions (16024 um) you can not do better without a
smaller sensor element which then has more noise and needs an
even stronger AA filter. Vicious circle at that point I guess,
so the D60 and later class CMOS sensors may be as good as
they are going to get unless Canon decides to allow some
Moire to quiet the crowds. Interesting trade off!

Bob Watt
 
Ray Amos said:
view at 100% and make a print to see if there really is any Moire'.
Click to expand...
Yes, only 100% shows the real moire. Downsized images can show a moire that is not real.

Uwe
Ray Amos said:
When I asked the Canon Rep about the Moire' problem I had with my
1D, he explained that there was greater than a 90% chance it would
not show up on the print. Mostly it was only on the monitor and is
only evident at some resolutions. He said if you view the image at
100 %, Moire' would not show up even on the monitor. He
demonstrated this on his laptop at different resolutions.

One of the reasons I sold my 1D was the Moire' problem with brick
buildings. I assumed since it was evident on my monitor, it would
be evident on the final print. I didn't even bother to print the
images.

To those of you who think you have a problem with Moire', I
recommend you try the above: view at 100% and make a print to see
if there really is any Moire'.

Ray Amos -
1v (waiting for my 1D2) and EF lenses from 17-500mm.
Click to expand...
 
Patrick Pai said:
Jim,

Reading your first post, I think you may be misunderstanding a
point I made. The designer will trade off moire with softness,
meaning he will adjust the strength of the AA filter to get more of
one or the other. I was not saying that he will decide whether or
not to have an AA filter. Since moire (or aliasing) is an effect
that is far more destructive, I believe that the prudent and
conservative decision is to lean towards softness, which can be
compensated once digitized, either in-camera or in the PC.
Ideally, the engineer should find the perfect bandwidth of a steep
(and expensive) AA filter so that it is visibly sharp and moire
effects are rare (the "sweet spot" you mention).
Click to expand...
Sorry Patrick,

I read your original post and it seemed to me that you were saying that you perferred a softer image to moire.

My reply was to point out that sharpening in the real world is rarely easy. High ISO images are problematic. As are images that are unintentionally soft from motion blur (slow shutter speed). This is also a problem when you are also shooting wide open at f2 or f2.8. Combine all three factors in a PJ environment, and you get low and high frequencies. So which do you sharpen?

Bear in mind that portraits are harder than they look. Without makeup, faces have all kinds of irregular features. Hair have different reflectivity and sizes. Eyeglasses will sharpen differently from the face.

So, it may be that moire is destructive to capture fidelity. (You should pop along to the Kodak and Fuji forums to see how they cope) But, there are additional issues.

That is why it is difficult for camera makers to choose a degree of compromise. Besides technical issues, Canon etc has to make a marketing decision.

That is why so many 1D users are unhappy with the MK2. They were comfortable with the image strategy. Moire wasn't a problem with the 1D. Suddenly, it's been changed. The technical issues are irrelevant. It's been done. The question is how to cope with the changed image strategy.

--
Jim
 
Dean Brown said:
Razor sharp but with moire. Right?

Isn't that why there is some issue with the 1Ds and moire? The AA
filter is "weaker"?
Click to expand...
Dean,

As I understand it, the Kodak doesn't come with an AA filter in the standard package. That is one reason it is so much cheaper.

Here's a link that discusses the issue of moire, and describes what you can do about it.

http://support.nikontech.com/cgi-bin/nikonusa.cfg/php/enduser/std_adp.php?p_sid=LDjdRD7h&p_lva=&p_faqid=9544&p_created=1080656943&p_sp=cF9zcmNoPSZwX2dyaWRzb3J0PSZwX3Jvd19jbnQ9NDcmcF9wcm9kX2x2bDE9NDImcF9wcm9kX2x2bDI9NDUmcF9jYXRfbHZsMT1_YW55fiZwX3BhZ2U9MQ**&p_li=

--
Jim
 
Jim,

No problem. I see your point now about the difficulties on fine tuning in real life. One thing I'd like to look at is characterizing the AA filter in the 1D II, and coming up with an inverting filter (in photoshop or in software) which should get you very close to an image with no AA filter and no moire. The best inverting filter depends on the amount ISO noise and it's spectral shape in the image. I realize that this is not a trivial solution, but I've spent a few years in the past working on a very similar problem in hard disk drives. Anyway, I'd have fun giving it a shot. If I am at all successful, you'll be sure to hear from me here.

Pat
Jim T H said:
Sorry Patrick,

I read your original post and it seemed to me that you were saying
that you perferred a softer image to moire.

My reply was to point out that sharpening in the real world is
rarely easy. High ISO images are problematic. As are images that
are unintentionally soft from motion blur (slow shutter speed).
This is also a problem when you are also shooting wide open at f2
or f2.8. Combine all three factors in a PJ environment, and you get
low and high frequencies. So which do you sharpen?

Bear in mind that portraits are harder than they look. Without
makeup, faces have all kinds of irregular features. Hair have
different reflectivity and sizes. Eyeglasses will sharpen
differently from the face.

So, it may be that moire is destructive to capture fidelity. (You
should pop along to the Kodak and Fuji forums to see how they cope)
But, there are additional issues.

That is why it is difficult for camera makers to choose a degree of
compromise. Besides technical issues, Canon etc has to make a
marketing decision.

That is why so many 1D users are unhappy with the MK2. They were
comfortable with the image strategy. Moire wasn't a problem with
the 1D. Suddenly, it's been changed. The technical issues are
irrelevant. It's been done. The question is how to cope with the
changed image strategy.

--
Jim
Click to expand...
 
Hi,

These days, it's called experience. ;) One gets to know these things when faced with them often enough. There are only a few major fabric weave categories that cause issues.

You are correct. The on-camera LCD won't properly tell the story. Sometimes it shows aliasing that isn't there at full size. Other times it won't show what is there, and that's a lousy surprise. :(

The 460 had no LCD, so that problem was easily avoided. You shot like you had film and then processed them all later. Thre was no choice of file type. You had only Raw, and had to use a plug-in with Photoshop to read the files. So, you had to process each shot one-by-one. Until you did, you didn't know if things were good or fouled up.

In the studio, you could hook the camera up to a computer via a SCSI port. Then, you could control the camera from the keyboard and see each shot as it came in. This was the best way to learn what filters to use for what fabric.

Now that we have filters in the cameras, it helps but does not mean that there won't be conditions that cause trouble. If you encounter trouble, you will have to do the same thing and use a softening filter, and you'll have to learn which one is the right one in the same manner.

Stan
Forrest said:
Are you going by the idea that if you can't see it on the camera's
LCD it won't be a distraction in print, or are you taking extra
shots just in case...?
Click to expand...
--
Amateur Photographer
Professional Electronics Development Engineer

Once you start down the DSLR path, forever will it dominate your destiny! Consume your bank account, it will! Like it did mine! :)

More info and list of gear is in my Posters' Profile.
 
You must log in or register to reply here.

Keyboard shortcuts

f
Forum
m
My threads
Mobile site
About
Editorial content
Cameras & Lenses
Community
All content, design, and layout are Copyright © 1998–2025 Digital Photography Review All Rights Reserved.
Reproduction in whole or part in any form or medium without specific written permission is prohibited.
When you use DPReview links to buy products, the site may earn a commission.
©GPS Media - Guides, Products, Services.
Back
Top