Blue Pattern Moire - Questions

[technoid]

Moiré is spatial frequency aliasing. As such it is most sensitive when using high quality lens, precisely focused, minimal diffraction (low F-stop), either red or, slightly worse, blue patterns, together with a pattern which repeats closely to the associated pixel spacing. Blue is the most vulnerable since it has the shortest wavelength and hence the smallest diffraction (Airy disk diameter).

Raw conversion algorithms can do a decent job with some types of moiré, especially pattern induced, localized color artifacts, but not others such as low frequency moire which can't be distinguished from an actual luminance ripple at the alias frequency.

Low pass filters, which are on most cameras do a reasonable job of mitigating this at the cost of lower resolution but by forcing a lower bound on resolution and decreasing resolution in general.

I recently, in looking at the tradeoffs and going through a quick bit of math chose the 50MP 5DSR. First new digital in quite a while. Last one was a 1DsIII.

I would occasionally have issues with moiré on repeating patterns or fabric and an F-stop of F8 or less, rarely with F11.

Because of the smaller pixel spacing on the 5DSR I expect moiré sensitivity to drop about 1 F-stop and that seems to have been the case though it is quite early.

So, with that as background, are there test charts specifically for quantifying moiré in the blue or red range where digital imaging is most sensitive to aliasing due to the 2x pixel spacing?

Also, almost all aliasing could be eliminated by using a raw converter that takes two images, one of which is very slightly out of focus. At least for stationary, aligned shots.

EtoA; I haven't seen a raw converter with that option but it would be relatively simple.

 

[J A C S]

So, with that as background, are there test charts specifically for quantifying moiré in the blue or red range where digital imaging is most sensitive to aliasing due to the 2x pixel spacing?

You can do that yourself. Shoot a B&W chart in daylight, and then in reddish light.

Also, almost all aliasing could be eliminated by using a raw converter that takes two images, one of which is very slightly out of focus. At least for stationary, aligned shots.

Or slightly misaligned. Photo Acute can do something similar. The pixel shift technology reduces aliasing, too. Photo Acute creates a DNG out of the RAWs. I have not tested it for moire though but it should help.

EtoA; I haven't seen a raw converter with that option but it would be relatively simple.

 

[technoid]

So, with that as background, are there test charts specifically for quantifying moiré in the blue or red range where digital imaging is most sensitive to aliasing due to the 2x pixel spacing?

You can do that yourself. Shoot a B&W chart in daylight, and then in reddish light.

Also, almost all aliasing could be eliminated by using a raw converter that takes two images, one of which is very slightly out of focus. At least for stationary, aligned shots.

Or slightly misaligned. Photo Acute can do something similar. The pixel shift technology reduces aliasing, too. Photo Acute creates a DNG out of the RAWs. I have not tested it for moire though but it should help.

EtoA; I haven't seen a raw converter with that option but it would be relatively simple.

Interesting. I'll have to take a look at it.

I just ran some Moiré tests with a blue, circular chart comparing my old 1DsIII with the 5DSR using the current 70-200 2.8L at 70. Adobe ACR, sharpening at 0.

I ran from F3.5 to F16.

Moiré was much stronger on the 5DSR below F8. At F11 the 5DSR was roughly the same as the 1DsIII at F16 but remained significant. At F16 Moiré was close to non-existent of the 5DSR.

One rather curious difference is that the 5DSR exhibited rather strange and pronounced color shifts into the yellow on some patterns at the lower F-stops. I suspect this may be due to a weaker CFA color separation (significant overlap of the color filters) in order to capture more photons.

This is, of course, a worst case test because the blue pixels are spaced 2x unlike the green and the blue also has the shortest wavelength so a higher F-stop is required to produce a low pass effect from diffraction. Interesting that even at F16 the 1DsIII, which has an AA filter, produced significant Moiré.

 

[technoid]

Blue Pattern Moire from F3.5 to F16, Top:5DS R, Bottom: 1DsIII

These are from a tiny section of a large image. The 1DsIII was taken at about 40', the 5DS R at 70'. 70-200L at 70.

Key points:

1. Moire is worse on the 5DS R at F-stops below F11. SDS R better at F16

2. The 5DS R Moire produces a large color shift.