Thanks for the feedback, and good job with the debanding luc_october. It's effective; however, you'll notice that fine vertical lines in the building on the bottom right have also disappeared. This is generally why I don't like noise removal... it lowers MTF, and when I go to print, I always choose the one without noise removal... kind of like how film grain sharpens prints.
Out of curiosity, what software did you use luc_october?
For those still looking for the banding, view at 100%, and look for vertical striations. Move the image around to see it easily.
Those saying a client won't see it-- maybe. But remember, this sky was actually not even that dark. The original exposure is 1" at f/16, ISO 200. It was, on average, RGB of (24,29,44) on a 0-255 scale in the original RAW. For the more technically inclined, the RAW 14-bit RGBG data is somewhere around (after a subtraction of 2047, the offset bias):
The read noise on this scale is 6 (as determined from dark frames). So we're well
above the noise floor. This just confirms what someone else here (or on photo.net, I forget) pointed out: that the FPN is overlayed over everything
, not just low signals. It's just more easily seen over low signals b/c of the percentage of the signal the FPN makes up. But that doesn't mean it won't be apparent elsewhere, especially in areas of smooth tonality.
Here's another example; original image on right, vignetting correction + a gentle shadow lift on left (+100 blacks, +22 shadows in LR4; +100 blacks is not as drastic as it sounds
, if you've used LR4 you know this):
View it at 100% here: http://cl.ly/GS7R/5DIIIBandingUponVignettingCorrection.jpg
Again, vertical banding is apparent even in image areas that aren't that dark to begin with. This was shot w/ the 35/1.4, so vignetting correction is often necessary in the corners. Or not, depending on your artistic intent.
My point is simply that there's no trace of this in any Nikon file I've seen thus far, so it's disappointing to me, & I would've expected better in 2012 from Canon.
I averaged 5 exposures to get rid of noise in the buildings (both shot & read noise is reduced by image averaging)... there was a good deal of noise there since the buildings were rather underexposed in order to not blow any channels in the moon. Rather than HDR this, I just averaged 5 exposures which was pretty effective in getting rid of noise in the buildings. The FPN isn't so apparent in the buildings b/c of all the other detail there (not an area of smooth tonality).
I performed masked sharpening so as to not sharpen the sky as much as everything else. Unmasked sharpening accentuated the FPN even more.
Image averaging is another approach to HDR -- you're effectively removing shadow noise so you can boost it more effectively. A camera with higher DR (lower shadow noise) would do better, of course, since the read noise contribution would be considerably lower (e.g. on a D800). Shot noise down there would still remain though, & so the image might still benefit from image averaging. But anyway, I digress at this point...