Aliasing and supersampling, why small pixels are good?

[ErikKaffehr]

Hi,

It may be that I have an obsession with aliasing, but anyway here is an article illustrating the issue:

http://echophoto.dnsalias.net/ekr/index.php/photoarticles/78-aliasing-and-supersampling-why-small-pixels-are-good

I am posting this topic here, because it is mostly an MFD issue.

The sample below is from that article:

Full size: http://echophoto.dnsalias.net/ekr/Articles/Aliasing2/Dollars.jpg

Best regards
Erik

[bjanes]

Hi,

It may be that I have an obsession with aliasing, but anyway here is an article illustrating the issue:

http://echophoto.dnsalias.net/ekr/index.php/photoarticles/78-aliasing-and-supersampling-why-small-pixels-are-good

I am posting this topic here, because it is mostly an MFD issue.

Best regards
Erik

Erik, that is an excellent article on aliasing. Presumably, the aliasing would be reduced with newer MFDBs with higher resolution.

Incidentally, your feather and real world images do not appear when the article is viewed with the latest versions of Firefox or Chrome on my Windows 7 Laptop. They do appear with the latest version of Internet Explorer on the laptop and with Safari on my iPad. ?HTML problem.

Regards,

Bill

[ErikKaffehr]

Bill,

Thanks for comments! 

I posted some of the images as TIFFs as I have seen that JPEGs seem to wash out some of the effects I wanted to show. I will convert all TIFFs to PNG.

Yes, smaller pixels on newer MFDs will reduce the problems, but they are often used with better lenses which enhances the problem. Anyway I would argue that increasing resolution is a good way to reduce aliasing.

Best regards
Erik

Erik, that is an excellent article on aliasing. Presumably, the aliasing would be reduced with newer MFDBs with higher resolution.

Incidentally, your feather and real world images do not appear when the article is viewed with the latest versions of Firefox or Chrome on my Windows 7 Laptop. They do appear with the latest version of Internet Explorer on the laptop and with Safari on my iPad. ?HTML problem.

Regards,

Bill

[Bart_van_der_Wolf]

Presumably, the aliasing would be reduced with newer MFDBs with higher resolution.

Hi Bill,

That would indeed help some, assuming the sensor array size stays the same and the sampling gets denser (smaller sensel pitch for the same image magnification factor).

Part of the effects shown, the false color artifacts, can be solved by different demosaicing and/or moiré suppression afterwards, but the origin of the artifacts is exacerbated by the combination of 'large' sensels relative to the level of image detail and the lack of an OLPF (which boosts the MTF near Nyquist).

As an alternative cure one could consider using a longer focal length (which magnifies the detail and reduces the risk of aliasing), and the use of stitching for a restoration of the FOV (stitching for resolution enhancement, which also boosts the overall MTF). Diffraction, by using a second exposure with narrower aperture, can help with postprocessing of the problematic areas within the DOF zone (image magnification differences become too large outside of the DOF zone).

Cheers,
Bart

[Fine_Art]

It seems to me that if the photographer chooses to use a sensor with no OLP, they have to make conscious effort to avoid taking pictures at ranges where the detail will make these artefacts.

I know the lenses I have used on Sony from A100 to A55 were capable of out resolving the sensors. Trouble was avoided by a well matched OLP. The lenses I have on the D600 are better according to DxO yet I still have no problems with the larger pixels, while not losing detail.

When I have used weak lenses on a sensor I found a tearing look to the detail. This may just have been an aberration like spherical or coma. Small pixels that try to go beyond the lens will probably have a similar look. It is not pretty. Many people will know what I mean from the corners of some of their shots.

I guess the obvious answer is no weak link is good. The whole system should be designed to a common level.

The weakest link is the resolution of the screens we have to shoehorn our images into.

[eronald]

defocus slightly, problem solved.

Edmund

It seems to me that if the photographer chooses to use a sensor with no OLP, they have to make conscious effort to avoid taking pictures at ranges where the detail will make these artefacts.

I know the lenses I have used on Sony from A100 to A55 were capable of out resolving the sensors. Trouble was avoided by a well matched OLP. The lenses I have on the D600 are better according to DxO yet I still have no problems with the larger pixels, while not losing detail.

When I have used weak lenses on a sensor I found a tearing look to the detail. This may just have been an aberration like spherical or coma. Small pixels that try to go beyond the lens will probably have a similar look. It is not pretty. Many people will know what I mean from the corners of some of their shots.

I guess the obvious answer is no weak link is good. The whole system should be designed to a common level.

The weakest link is the resolution of the screens we have to shoehorn our images into.

[alifatemi]

Bill, I have taken some pics with iQ260 with lots of aliasing with Schneider 80mm. In theory, higher resolution sensors reduce it but in practice it is not so.

[ErikKaffehr]

Hi Ali,

You get aliasing if the lens outresolves the sensor. I would rather say if the lens has significant MTF past the Nyquist limit. Anyway, good lens and good technique leads to aliasing. The question is if it is objectionable or not.

The Sony Alpha 77 used in these tests has 3.9 micron pixel pitch and probably also some OLP filtering, so it does show little aliasing.

Thanks for sharing your experience!

Best Regards
Erik

Bill, I have taken some pics with iQ260 with lots of aliasing with Schneider 80mm. In theory, higher resolution sensors reduce it but in practice it is not so.