Thoughts about perception or occurrence of aliasing

[Erik Kaffehr]

Hi,

I started a thread discussing aliasing over on the medium format threads: https://www.dpreview.com/forums/post/67453050

An interesting observation about the discussion on the issue is that it seems that few observers really note aliasing in their images, while some have many issues with it, including myself.

So, I am asking, what is the reason that aliasing is not a problem for many users of cameras without OLP (Optical Low Pass) filters.

I have come up with some explanations.

I think that photographers often appreciate crispness, but don't see as false detail, unless it is seen as color moiré.

But, I also think it may be subject dependent.

• If we consider textiles, which are very moiré prone, they are often shot at short distance, where DoF is very limited. So, we may seen moiré but in a very narrow part of the image.
• I would also think that the 'pitch' of the textile may be quite coarse when projected on the sensor, so high frequency components may have low amplitude.
• Stopping down increases DoF, but once we are past f/8 diffraction may reduce high frequency content.

Shooting at long distance the situation is different:

• Large parts of the image will be in near ideal focus.
• Image detail projected on the sensor will have very small detail.
• Shooting architecture or landscape, we often have very controlled workflow, focusing very carefully, using near optimal apertures and keeping vibrations down.

If we look at say medium format architecture images, we often see aliasing artifacts, but they may not be very obvious.

What I also see is that we have kind of an eternal race. We want better lenses, but those lenses need better sensors to resolve the detail delivered by the lenses. But, that detail will not be crispy, so we want better lenses...

Best regards

Erik

 

[hjulenissen]

Hi,

I started a thread discussing aliasing over on the medium format threads: https://www.dpreview.com/forums/post/67453050

An interesting observation about the discussion on the issue is that it seems that few observers really note aliasing in their images, while some have many issues with it, including myself.

So, I am asking, what is the reason that aliasing is not a problem for many users of cameras without OLP (Optical Low Pass) filters.

I have come up with some explanations.

I think that photographers often appreciate crispness, but don't see as false detail, unless it is seen as color moiré.

But, I also think it may be subject dependent.

• If we consider textiles, which are very moiré prone, they are often shot at short distance, where DoF is very limited. So, we may seen moiré but in a very narrow part of the image.
• I would also think that the 'pitch' of the textile may be quite coarse when projected on the sensor, so high frequency components may have low amplitude.
• Stopping down increases DoF, but once we are past f/8 diffraction may reduce high frequency content.

Shooting at long distance the situation is different:

• Large parts of the image will be in near ideal focus.
• Image detail projected on the sensor will have very small detail.
• Shooting architecture or landscape, we often have very controlled workflow, focusing very carefully, using near optimal apertures and keeping vibrations down.

If we look at say medium format architecture images, we often see aliasing artifacts, but they may not be very obvious.

What I also see is that we have kind of an eternal race. We want better lenses, but those lenses need better sensors to resolve the detail delivered by the lenses. But, that detail will not be crispy, so we want better lenses...

Best regards

Erik

Ad perception:

Aliasing seems to be more objectionable for periodic structures with a periodicity that slowly "sweeps" in and out of the nyquist frequency of the spatial sampling process. Like a fabric that excite obvious false frequencies some places, but not others.

Also, for sharp edges (electricity poles) that are close to a single pixel wide and slightly tilted, if you are doing really low resolution (or zooming in a lot).

For "chaotic" natural content, like foliage or sand or similar, it seems a lot less annoying.

Ad dsp:

It seems sensible from an engineering PoV to maximise the general spatial bandwidth (sharp lense, many megapixels), then assure that you get a comprehensible signal within that bandwidth (OLPF).

 

[Tom Axford]

Several years ago I did some experiments with a piece of text to see the effect of aliasing on the legibility of the text in comparison to an image with strong AA. This was simulated in software rather than done with a real camera.




Image of some text with point sampling (which gives the most extreme aliasing)




Image of the same text with strong AA filtering before sampling

The first image clearly looks much sharper, but the bottom line of text is not legible.

The second image was blurred with a gaussian blur filter before sampling. It looks much less sharp but the bottom line of text can be just about read (with a little guesswork).

Many photographers today seem to prefer very sharp images (even if most of the sharpness is made up of false detail). Personally, I prefer to see less sharp detail, but know that what is present in the image is as true a representation as possible. I really don't like the idea of false detail.

 

[Horshack]

If we look at say medium format architecture images, we often see aliasing artifacts, but they may not be very obvious.

I think most don't have an issue with aliasing because it's not obvious to them in many photographs. If one mistakes aliasing for detail then the image just looks sharper.

The pixel shifting feature on many current cameras provides a method to actually see which detail is aliased, at least the aliasing aggravated by the bayer arrangement. For example here's a comparison of a 24MP FF Nikon Df single exposure vs a 4-shot pixel shift, where each pixel is sampled by all four 2x2 bayer filters:

Animation: Nikon Df single exposure vs 4-shot pixel shift, 300% nearest-neighbor crop

 

[Ephemeris]

If we look at say medium format architecture images, we often see aliasing artifacts, but they may not be very obvious.

I think most don't have an issue with aliasing because it's not obvious to them in many photographs. If one mistakes aliasing for detail then the image just looks sharper.

The pixel shifting feature on many current cameras provides a method to actually see which detail is aliased, at least the aliasing aggravated by the bayer arrangement. For example here's a comparison of a 24MP FF Nikon Df single exposure vs a 4-shot pixel shift, where each pixel is sampled by all four 2x2 bayer filters:

Animation: Nikon Df single exposure vs 4-shot pixel shift, 300% nearest-neighbor crop

That's a helpful animation thanks.

 

[sharkmelley]

So, I am asking, what is the reason that aliasing is not a problem for many users of cameras without OLP (Optical Low Pass) filters.

Maybe it's lens quality. A lens needs be capable of resolving very fine detail before aliasing becomes a problem.

 

[John Vickers]

So, I am asking, what is the reason that aliasing is not a problem for many users of cameras without OLP (Optical Low Pass) filters.

Maybe it's lens quality. A lens needs be capable of resolving very fine detail before aliasing becomes a problem.

Not in my experience:

Testing with Siemens star targets, I find I need to stop down to f/16-f/22 to reduce aliasing to an acceptable (to me) level on a 24MP FF Bayer camera, and f/11-f/16 on a 47MP FF Bayer camera.

Good modern FF lenses tend to be sharpest at around f/3-f/4.

Even a budget FF kit zoom is likely to be diffraction limited by f/8.

[Edit:] Both the cameras I referred to above lacked any OLPF.

 

[Tom Axford]

An interesting observation about the discussion on the issue is that it seems that few observers really note aliasing in their images, while some have many issues with it, including myself.

Moire is often much more objectionable in movies than in still photographs, as is noted in this article.

 

[J A C S]

Moire is just one manifestation of aliasing.