Is sharper always better?

There have been various discussions about the merits or otherwise of AA filters on cameras, such as in this thread.

The following example is computer generated and not taken from real cameras in which the different colours of different pixels complicate the matter hugely. Instead, this example assumes the camera is monochrome and that the lens is perfectly sharp (which, of course, is never true in practice). It also assumes that the pixels are small compared to the spacing between pixels. So the result is more extreme than is the case with real cameras, but it does serve to illustrate the fact that sometimes the blurrier image from a camera with an AA filter can be "better" than the sharpest possible image.




The image with no AA filter - it looks very crisp and sharp. Can you read the bottom line of text?




The same image with a strong AA filter - it looks much less crisp and sharp, but despite that the text is more pleasing to look at, if you are trying to read it. Can you read the bottom line of text?

Of course, many people prefer the crisper look of the first image and they are not photographing documents with lots of text in them. Hence the popularity of cameras without AA filters.

You can also use the image comparison tool here and focus on the text.

Panasonic G85 vs G7 , the G85 has no AA filter and seems easier to read the text than

the AA filtered G7.

Tom Axford said:

There have been various discussions about the merits or otherwise of AA filters on cameras, such as in this thread.

The following example is computer generated and not taken from real cameras in which the different colours of different pixels complicate the matter hugely. Instead, this example assumes the camera is monochrome and that the lens is perfectly sharp (which, of course, is never true in practice). It also assumes that the pixels are small compared to the spacing between pixels

Click to expand...

This is a pretty odd assumption. The effect of fill Factor on most sensors is extremely high approaching 100% (micro lenses, BSI). Effectively this part of your sample is not representative of reality for a few different reasons.

Tom Axford said:

. So the result is more extreme than is the case with real cameras, but it does serve to illustrate the fact that sometimes the blurrier image from a camera with an AA filter can be "better" than the sharpest possible image.

Click to expand...

If you wish to show a more practical example of the effect of AA filters, look at the camera comparison widget with the Canon 5D s and 5dsr, or the Nikon D800 and d800e. At least these would give an actual and not a synthetic comparison and would be eminently fair since it would show the effect of otherwise identical sensors.

Tom Axford said:

The image with no AA filter - it looks very crisp and sharp. Can you read the bottom line of text?


The same image with a strong AA filter - it looks much less crisp and sharp, but despite that the text is more pleasing to look at, if you are trying to read it. Can you read the bottom line of text?

Of course, many people prefer the crisper look of the first image and they are not photographing documents with lots of text in them. Hence the popularity of cameras without AA filters.

Click to expand...

I doubt that your samples would/would not encourage the popularity with/ without AA filters.

To answer your question, I generally shoot a lot of subjects with fabric as in wardrobe, and generally prefer a camera with an AA filter. It is have some interest that I have more moire with my cameras with an AA filter than those without, however the cameras with an AA filter do have a tendency to minimize the artifact. It is always subject dependent but I do prefer sharpness and chrispness.

--
Charles Darwin: "ignorance more frequently begets confidence than does knowledge."
tony
http://www.tphoto.ca

Or for that matter any woman over about 12

Their criteria may not be ours

--
"Good judgement is generally the result of experience. Experience, unfortunately, is generally the result of bad judgement" - attributed to a whole bunch of folks

Tom Axford said:

There have been various discussions about the merits or otherwise of AA filters on cameras, such as in this thread.

The following example is computer generated and not taken from real cameras in which the different colours of different pixels complicate the matter hugely.

Click to expand...

No explanation about how those computer generated examples were produced?

No way to confirm how close those simulations come to actually representing what they're supposed to represent?

No photographic comparison of a real AA filter vs. no AA filter?

I'm not sure that was helpful. We can always find examples online showing the good and bad of AA filters:

I prefer not dealing with aliasing artifacts. Running the side with the AA filter through basic sharpening and now comparing it to the Side without the AA...one sees detail virtually identical, without the aliasing artifacts. The benefits to detail and acutance by eliminating the AA filter are going to be invisible in any sized prints I can think of.

No AA filter









With AA and sharpened

The OP looks at Anti-Aliasing Filters versus none with larger sensors in medium-cost cameras, using text examples.

In my perception, scanning text documents avoids a camera altogether. I constantly scan purchasing tickets for the accountant. Doing this I classify as 'Documentation' rather than as 'Photography'.

In 'Photography' I deal with the small-sensor pocket zooms I can afford, and the question becomes 'How much in-camera sharpening do I have to put up with?' Sometimes this matters when I do wall prints later.

This is a camera size picture from the Olympus SH-1:


At 200% the halo around the sky line becomes noticeable. I did the stills on the SH-1 because the SH-2 did a video for a time-lapse video, so I did not want to touch it.

This is a similar SH-2 picture, taken at the same 400 ISO as the SH-1 picture:


The halo is still there, but less noticeable. The SH-2 has RAW, and the firmware writers decided on less sharpening, and consequently on less noise reduction.

Up to full screen presentation, the difference hardly shows. But I do 33x23 inch prints from pocket zooms, and a magnifier on the print will emphasize the halo.

My apologies for my distraction from the original intent in this thread.

Henry

--
Henry Falkner - SH-2, SH-1, SH-50, SP-570UZ

If it were truly sharper, there wouldn't be gaps in the text lines. So I don't think this example quite illustrates the point. AA filters blur, they don't smooth. What you're illustrating is an example of filling/smoothing as in screen smoothing.

CarlosMP said:

You can also use the image comparison tool here and focus on the text.

Panasonic G85 vs G7 , the G85 has no AA filter and seems easier to read the text than

the AA filtered G7.

Click to expand...

Does anyone know why Panasonic decided to use a AA filter for the G7 camera. It seems strange as it's not used for the G85 camera.

Brian

sybersitizen said:

Tom Axford said:

There have been various discussions about the merits or otherwise of AA filters on cameras, such as in this thread.

The following example is computer generated and not taken from real cameras in which the different colours of different pixels complicate the matter hugely.

Click to expand...

No explanation about how those computer generated examples were produced?

No way to confirm how close those simulations come to actually representing what they're supposed to represent?

No photographic comparison of a real AA filter vs. no AA filter?

I'm not sure that was helpful. We can always find examples online showing the good and bad of AA filters:

Click to expand...

Are the photos on the left taken with the AA filter?

Brian

Tom Axford said:

Of course, many people prefer the crisper look of the first image and they are not photographing documents with lots of text in them. Hence the popularity of cameras without AA filters.

Click to expand...

Here's two shots through the same lens on two different sensors. (Sigma 55-200 at 55mm f/4)

Canon SL1 16MP with AA filter...



Fuji X-T2 24MP without any AA filter...



Not a direct comparison I realise with the different MP, but it does show the small increase in sharpness I had with my newer camera.

It does show with my astro nightscapes.

The Canon would always have slightly softer looking stars as if they were always a tiny bit out of focus...



The Fuji gives more pinpoint stars and you can see more of a size difference between brighter stars and the dimmer stars...



All in all my keeper rate has been much higher with much better results...



--
https://www.flickr.com/gp/158098310@N03/ja5n6b

tony field said:

Tom Axford said:

There have been various discussions about the merits or otherwise of AA filters on cameras, such as in this thread.

The following example is computer generated and not taken from real cameras in which the different colours of different pixels complicate the matter hugely. Instead, this example assumes the camera is monochrome and that the lens is perfectly sharp (which, of course, is never true in practice). It also assumes that the pixels are small compared to the spacing between pixels

Click to expand...

This is a pretty odd assumption. The effect of fill Factor on most sensors is extremely high approaching 100% (micro lenses, BSI). Effectively this part of your sample is not representative of reality for a few different reasons.

Click to expand...

I am sorry that it wasn't clear from my OP that these examples were not intended to match real cameras. They are extreme example of aliasing v. anti-aliasing, so that the differences are very obvious. No real cameras will produce images as extreme as these.

But I hope they do illustrate why scientists, engineers and many others working in the field of imaging regard aliasing as something to be avoided in most circumstances.

sybersitizen said:

Tom Axford said:

There have been various discussions about the merits or otherwise of AA filters on cameras, such as in this thread.

The following example is computer generated and not taken from real cameras in which the different colours of different pixels complicate the matter hugely.

Click to expand...

No explanation about how those computer generated examples were produced?

Click to expand...

The first image was downsampled with no interpolation between points. The second image was downsized in Lightroom using its standard method of downsizing. I don't know what AA algorithm Lightroom uses, but it is very effective.

The original image before downsizing is this:

sybersitizen said:

No way to confirm how close those simulations come to actually representing what they're supposed to represent?

Click to expand...

They are not supposed to represent anything other than themselves!

Some lenses are softer than others.
So, which gives better results, a sharp lens on a cam with AA filter
or a softer lens on a cam with no AA filter ?
The permutations are endless and will come down to personal choice,
usually of micro-contrast/grain/final sharpness preference.
I love the way I can lean into the sharpening on my old lenses on my A7R,
just more forgiving and "photo" like.
Talking about the AA filter in isolation won't produce a definitive answer for my 2 cents.

Schrodingers_cat said:

Or for that matter any woman over about 12

Their criteria may not be ours

Click to expand...

Certainly portraits should not be too sharp. That's why "portrait lenses" with plenty of spherical aberration were so popular.

I see that Leitz is re-issuing the Thambar lens at an absurdly high price. There will probably be imitators at lower prices.

But you don't need to go to such extremes. A legacy f/1.4 or f/1.2 lens is likely to be quite soft wide open, enough to reduce skin texture to acceptable levels.

I agree,

my em52 ( no AA) produces better skin texture detail then my em51 (AA) for macro aa less sensors are great wouldn't go back and i have never seen fabric moire on any dance school costumes.

Don

Brisn5757 said:

CarlosMP said:

You can also use the image comparison tool here and focus on the text.

Panasonic G85 vs G7 , the G85 has no AA filter and seems easier to read the text than

the AA filtered G7.

Click to expand...

Does anyone know why Panasonic decided to use a AA filter for the G7 camera. It seems strange as it's not used for the G85 camera.

Brian

Click to expand...

In fact as a norm to the industry, all Panasonic cameras (as well as majority of Olympus) have AA filter throughout the entire 12Mp era and also most of the 16Mp era until GX85, which is the first to use a newer hardware standard (floating sensor for a better 5-axis IBIS system, IBIS support video, new shutter shock free shutter, and also removal of the AA filter etc. IIRC EM10-II or EM1 of Olympus be the first to have AA filter removed). Since then, every Panasonic camera including G85, GH5, G9 etc uses a similar (of course with improvement over generation/class) hardware standard. G7 and GX8 be the last to use the old hardware standard (except for the entry class GFs like GX850 a.k.a. GF9 that still uses a fixed sensor).

Panasonic claimed that removal of the AA filter would expect 10% extra sharpness from older models have AA filter. But as per my observation between GX7 (has AA filter) and GX85 (no AA filter), the improvement in sharpness actually is quite obvious and should be more than the claimed 10%. Of course, beside removal of AA filter, the improved algorithm on GX85 also contributes to the improvement. Despite GX85's images come out sharper, they also looked more natural than those from GX7 which under close inspection, might feel sort of digital fingerprint...

Because of that extra sharpness, it makes the 16Mp non AA filter image be highly comparable to the 20Mp AA filter image (from GX8).

To me, sharper is always better .

--
Albert

Tom Axford said:

sybersitizen said:

Tom Axford said:

There have been various discussions about the merits or otherwise of AA filters on cameras, such as in this thread.

The following example is computer generated and not taken from real cameras in which the different colours of different pixels complicate the matter hugely.

Click to expand...

No explanation about how those computer generated examples were produced?

Click to expand...

The first image was downsampled with no interpolation between points. The second image was downsized in Lightroom using its standard method of downsizing. I don't know what AA algorithm Lightroom uses, but it is very effective.

The original image before downsizing is this:

Tom Axford said:

No way to confirm how close those simulations come to actually representing what they're supposed to represent?

Click to expand...

They are not supposed to represent anything other than themselves!

Click to expand...

You're being intellectually dishonest or disingenuous, whatever word you put to it. Giving us a computer text aliasing/antialiasing example which has nothing to do with photography. So - what was gained by posting it?

??

Ah, but the colors are nicer with the Canon.