Moiré Removal Algorithms

[Chris Kern]

Michael's piece today on the Nikon D800/D800E got me wondering . . .

I wonder if one of the more technically sophisticated forum participants could explain in general terms—I have no background in digital signal processing—what algorithms are used in post-processing software such as Lightroom, Capture One and Capture NX to reduce moiré.

I've done a little websurfing and to the extent that I've understood what I read (probably >not and <well) it seems the two more common techniques are (1) to simply paint over the pattern with an appropriately-colored sample from elsewhere in the image (e.g., an area of the same fabric that doesn't exhibit visible moiré) and (2) to replace the red and blue luminance data with data from the green channel (because there are more green sensels in a Bayer matrix, and therefore they are less likely to collect false detail from repetitive patterns).

Is that how it's done by current software products that offer moiré-reduction tools?  Are there other techniques besides the two I've described above?  Assuming I've described them correctly, of course.

Chris

[Schewe]

Moiré removal algorithms vary a lot, but the basic method of removing color moiré is to reduce the color saturation and blur the color. It's more complicated than that of course. I don't know exactly what C1 and LR are doing, but I think it's basically a color reduction/color blur locally. It does "ok" on pure luminance moiré which is much harder to remove.

[Chris Kern]

Moiré removal algorithms vary a lot, but the basic method of removing color moiré is to reduce the color saturation and blur the color. [ . . . ] It does "ok" on pure luminance moiré which is much harder to remove.

Thanks, Jeff.  Would I be correct in inferring from what you said—I'm still feeling my way here on uncertain ground—that these post-processing products reduce moiré without adversely affecting capture resolution?  In other words, that they may introduce slightly erroneous colors (whatever that means; my eye and brain no doubt interpret color differently than yours) and modify the overall luminance of the area harboring the interference pattern, but that detail acquired when the shutter was snapped would not inherently be likely to be damaged by using these algorithmic methods to reduce moiré.  Is that correct?

Finally, and slightly off-topic, do you agree with Michael's conclusions that moiré effects with a camera such as Nikon's D800E are likely to be (1) rare and (2) easily corrected?

Chris

[Schewe]

...that these post-processing products reduce moiré without adversely affecting capture resolution?

Well, of course the moiré will impact color and sharpness...the fact that some low level blurring is going on means, well, that some blurring is going on, ya know?

As for whether or not the D800E moiré will be rare...I don't know. My experience with moiré is it's really hard to find with high rez sensors...really, I've tried to find example to help the ACR/LR engineers when they were working on the moiré brush. I couldn't find any really good examples in my library. Then for some reason I found a perfect example (that I used the LR4 videos) where I found moiré on an IQ 180 shot. Sure enough, it was the subject frequency vs the sensor frequency and I saw moiré when zoomed in. The LR4 moiré fixed it fine...again, read Mike's statement of when you might or might not moiré based on what you are shooting. If you are shooting mechanical patterns or fabric, it's much more likely than in landscapes...which is what he shoots mostly.