Reading mode: Light Dark

Thoughts on the vertical banding in some SD1M images

Go to last post
This is old news. Yes, bandng is in the raw data See my post from May.

forums/read.asp?forum=1027&message=41420931
 
Hi Maple,

Took a look at some of your images....beautiful places, great shots. Manukau-Head-072 (with the 17-70's remarkable sharpness stopped down a bit, as we've both been saying), was one of my favorites. But the sky....

Anyway, I did a quick edit, and my approach to the sky turned out better than expected.

I first selected the dark foreground hill on the left (quick selection tool) and gave it Shadows 9 in PS to bring up more detail.

Then began selecting the sky from the bottom (magic wand with tolerance 8 to avoid features on the horizon--it's fiddly, keep adding to it till you get the whole sky)

I then did a couple of filters (two 'blur more's in succession). Not much help. But I next tried a couple of smart blurs in succession (radious 20, threshold 25, quality low, mode normal) And this appears to have cleaned up the banding without messing up the sky's gradations in tone and hue. Give it a try and tell me what you think.

Craig
 
Yes, you have done a very thorough analysis already. I read it but in a haste, so did not get a firm grip of the stuff.

Again, thank you for the most helpful link.
--
Maple
 
Thank you for that link - you have made a more thorough test than me.
And - your conclusions are compatible with mine.
The vertical stripes in John's image was less pronounced though.

Interesting that the stripe pattern is consistent from image to image. This gives some possibility to make a tool that removes the pattern.

I have also seen posterization banding with enhanced edges in skies, i.e. where the level is following a sawtooth curve. None such in John's image - so that might have been an effect of post processing. But ... if I can find those images again and can get a RAW version, then I can check.

--
Roland

X3F tools:
http://www.proxel.se/x3f.html
https://github.com/rolkar/x3f
 
From what I have seen I believe it would be possible to effectively remove the fixed pattern noise that shows up in skies and it really should be done before conversion from Raw format. Whats of more concern to me now is the vertical banding noise that shows up when the red channel is underexposed. This seem to be related to the fact that the new sensor is most sensitive in blue and least in red. This noise is variable from frame to frame and may be related to read and amplifier electronics. It looks like the dp2m may have addressed some of this banding noise issue but we will have to see some more challenging images from that camera.

Joe
Roland Karlsson said:
Thank you for that link - you have made a more thorough test than me.
And - your conclusions are compatible with mine.
The vertical stripes in John's image was less pronounced though.

Interesting that the stripe pattern is consistent from image to image. This gives some possibility to make a tool that removes the pattern.

I have also seen posterization banding with enhanced edges in skies, i.e. where the level is following a sawtooth curve. None such in John's image - so that might have been an effect of post processing. But ... if I can find those images again and can get a RAW version, then I can check.

--
Roland

X3F tools:
http://www.proxel.se/x3f.html
https://github.com/rolkar/x3f
Click to expand...
 
jbr said:
From what I have seen I believe it would be possible to effectively remove the fixed pattern noise that shows up in skies and it really should be done before conversion from Raw format. Whats of more concern to me now is the vertical banding noise that shows up when the red channel is underexposed. This seem to be related to the fact that the new sensor is most sensitive in blue and least in red. This noise is variable from frame to frame and may be related to read and amplifier electronics. It looks like the dp2m may have addressed some of this banding noise issue but we will have to see some more challenging images from that camera.
Click to expand...
Yes ... the new sensor has a thicker blue layer. The old sensor had HUGE problems with tungsten light. This is probably a fix for that.

And its a zero sum game. If you give something to the blue, you have to take from the green or red.

Changing the thickness of the layers affects the color conversion greatly. So ... the new sensor needs new algorithms (or at least new settings) there.

--
Roland

X3F tools:
http://www.proxel.se/x3f.html
https://github.com/rolkar/x3f
 
applying the filter sequence I suggested above on a (cloudless) sky, here are before and after crops from the sky of the Maple landscape I mentioned:





Again, this was a quick and dirty edit--I made no effort to refine it via alternative versions. But some may find it of use (again, the sky alone was selected in PS when the filters were applied
 
Craig Philips said:
applying the filter sequence I suggested above on a (cloudless) sky, here are before and after crops from the sky of the Maple landscape I mentioned:
Click to expand...
The mountains in the background is much fuzzier and the village/town also has less definition.
Craig Philips said:
Again, this was a quick and dirty edit--I made no effort to refine it via alternative versions. But some may find it of use (again, the sky alone was selected in PS when the filters were applied
Click to expand...
I assume a tailor made filter that knows aboyt the stripes and is applied on the RAW data should be able to remove the stripes without any ill effects.

Unfortunately I dont have any camera containing such sensor. If I had, I would probably have made such a filter.

--
Roland

X3F tools:
http://www.proxel.se/x3f.html
https://github.com/rolkar/x3f
 
Hi Roland

As you observed, my quick edit indeed soften the tops of the mountains. I don’t know whether PS applies a 2 pixel feather by default on magic wand selections, but I later tried specifying a 1 pixel “feather” after selecting the sky and the same fuzzy horizon persisted. However, I think you are imagining that the village/town lost definition—no filter was applied to it.

In return for a much improved sky I’d be inclined to call the blurred mountain range “aerial perspective” and live with it, but as it turned out there was a solution.

What do you think it was?



 
Roland Karlsson said:
I assume a tailor made filter that knows aboyt the stripes and is applied on the RAW data should be able to remove the stripes without any ill effects.

Unfortunately I dont have any camera containing such sensor. If I had, I would probably have made such a filter.

--
Click to expand...
In return for your banding filter
--
Maple
 
jbr said:
From what I have seen I believe it would be possible to effectively remove the fixed pattern noise that shows up in skies and it really should be done before conversion from Raw format. Whats of more concern to me now is the vertical banding noise that shows up when the red channel is underexposed. This seem to be related to the fact that the new sensor is most sensitive in blue and least in red. This noise is variable from frame to frame and may be related to read and amplifier electronics. It looks like the dp2m may have addressed some of this banding noise issue but we will have to see some more challenging images from that camera.
Click to expand...
Aha! I've seen banding in flat red areas from DP2M samples on flickr, this sounds like an explanation. But interesting if this will improve performance in low-indoor-light. Looking forward to test my future little Merrill.
 
Johan Borg said:
Aha! I've seen banding in flat red areas from DP2M samples on flickr, this sounds like an explanation. But interesting if this will improve performance in low-indoor-light. Looking forward to test my future little Merrill.
Click to expand...
Do you have a link to the pictures? I have seen some banding, but mostly not in the DP2M shots on flickr.

--
AF takes away the Zen in photography
Photos: http://www.flickr.com/photos/prebenr/
M42 on Foveon: http://www.flickr.com/groups/m42-foveon/
 
Craig Philips said:
Hi Roland

As you observed, my quick edit indeed soften the tops of the mountains. I don’t know whether PS applies a 2 pixel feather by default on magic wand selections, but I later tried specifying a 1 pixel “feather” after selecting the sky and the same fuzzy horizon persisted. However, I think you are imagining that the village/town lost definition—no filter was applied to it.
Click to expand...
OK - looking again I see that you are right, the town is not affected. Easy to imagine things :)
Craig Philips said:
In return for a much improved sky I’d be inclined to call the blurred mountain range “aerial perspective” and live with it, but as it turned out there was a solution.

What do you think it was?
Click to expand...
I have generally found that even if you have a sharp selection, you are affected by nearby areas, for blurring and other non pixel local manipulations. I think that is because the selection only applies to the output. The input is still taken from the entire image.

What I personally would have done there would be to decrease the selected area and accepting the stripes near to the horizon.

--
Roland

X3F tools:
http://www.proxel.se/x3f.html
https://github.com/rolkar/x3f
 
Probably won't help. The banding could be camera specific.
 
You must log in or register to reply here.

Keyboard shortcuts

f
Forum
m
My threads
Mobile site
About
Editorial content
Cameras & Lenses
Community
All content, design, and layout are Copyright © 1998–2025 Digital Photography Review All Rights Reserved.
Reproduction in whole or part in any form or medium without specific written permission is prohibited.
When you use DPReview links to buy products, the site may earn a commission.
©GPS Media - Guides, Products, Services.
Back
Top