sarangiman said:
And here I thought I knew a lot about the tech side of photography. I have ZERO idea what you just said, but you sound smart.
Upvote
0
5D3 Dynamic Range
- Thread starter jaayres20
- Start date
- Status
And here I thought I knew a lot about the tech side of photography. I have ZERO idea what you just said, but you sound smart.
Yet another example of FPN on my 5D Mark III...
Here's the full shot:
And here's a 100% crop:
See the full image at 100% here (drag it around on your screen to see banding): http://cl.ly/GSVA/Canon5DIII-FPNinSkies.png
I personally find this somewhat unacceptable in the light of EXMOR sensors, though I understand others may not.
FYI I averaged 5 exposures (this is ISO 200), which actually accentuates FPN b/c temporally variant noise is removed.
FYI the blue sky in the original RAW is 10% R, 17%G, 24%B... hardly unreasonable, IMHO, for a little shadow lift.
Yet it shows this kinda FPN.
This is sort of unacceptable to me in the face of technology like the D800. YMMV, of course, but I think Canon should step it up & offer something somewhat competitive.
Here's the full shot:
And here's a 100% crop:
See the full image at 100% here (drag it around on your screen to see banding): http://cl.ly/GSVA/Canon5DIII-FPNinSkies.png
I personally find this somewhat unacceptable in the light of EXMOR sensors, though I understand others may not.
FYI I averaged 5 exposures (this is ISO 200), which actually accentuates FPN b/c temporally variant noise is removed.
FYI the blue sky in the original RAW is 10% R, 17%G, 24%B... hardly unreasonable, IMHO, for a little shadow lift.
Yet it shows this kinda FPN.
This is sort of unacceptable to me in the face of technology like the D800. YMMV, of course, but I think Canon should step it up & offer something somewhat competitive.
Upvote
0
sarangiman said:
I must be blind. I don't see any fixed-pattern noise in that shot.
I DO see some clouds generally running horizontally through the scene. Perhaps that's what you're referring to?
b&
Upvote
0
TrumpetPower! said:
I see it. Its only there at 100% on a high end calibrated monitor. Would a client ever see it? Nope.
Upvote
0
P
PhilDrinkwater
Guest
Tcapp said:
I agree. I didn't see it on my work monitor, but I can see it on my "photo" monitor at home.
I think the client angle is one I would take. If my clients wouldn't notice it, it's not actually a problem unless I want to make it a problem.
Different clients have different expectations of course. A high end beauty campaign - yes they may notice and they wouldn't be that happy, especially if the retouch budget increased to fix it. A high street studio taking portraits? No.
Upvote
0
L
luc_october
Guest
FIXED 
http://dl.dropbox.com/u/16954130/Canon5DIII-FPNinSkiesdebanding.jpg
I recommend using debanding plugins in post.
http://dl.dropbox.com/u/16954130/Canon5DIII-FPNinSkiesdebanding.jpg
I recommend using debanding plugins in post.
Upvote
0
I have a few questions. The first is why did you blend 5 exposures. As you say, that tends to accentuate noise and I find that long exposure shots like this don't have that much dynamic range. In addition, digital sensors don't cope as well with film when exposing for several minutes (i.e. 15 or so minutes plus), I have seen that comment from Nikon shooters too. The second question is how much sharpening was performed. Again, sharpening tends to accentuate noise (or at least make it more visible) I tend not to sharpen, unless I intend to print, in which casem the reproduction isn't large enough to show any noise. The final questions are how long was the exposure and how much did you push a) the image overall and b) the shadows, as the EXIF info has been wiped.
Upvote
0
luc_october said:FIXED
http://dl.dropbox.com/u/16954130/Canon5DIII-FPNinSkiesdebanding.jpg
I recommend using debanding plugins in post.
what debanding plugin did you use?
Upvote
0
Hi folks,
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):
R: 70
G1:260
B: 290
G2: 260
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.
Kernauk:
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...
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):
R: 70
G1:260
B: 290
G2: 260
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.
Kernauk:
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...
Upvote
0
sarangiman said:
Right at zero though say a pixel is, on different frames, value -5,-1,0,1,-2,0,-1,0,-1,-2,-4,1,1,-1,2,0,2,6,-3,3,0 and you avg that you get a different value then if you chop off all the negative results and avg over 0,0,0,0,0,1,1,1,2,2,6,3, you end up with a bit higher number and on avg with a million frames it would not avg to 0 but to a bit above I'd think. Whether that really matters a lot I don't know since I haven't done that sort of photography.
Upvote
0
- Status
Similar threads
