GH2 What causes this color banding in the sky?

good samples and tests, thanks..
the sky trails footage is interesting.. seems like the banding 'went' to the other side (started on the left, went to the right).
i guess that if i'll try to avoid a lot of camera movement containing 'banded' areas, it will be easier to fix (opposed to panning across the sky like in the trailing footage, totally exposing the weakness)
 
roei z said:
the sky trails footage is interesting.. seems like the banding 'went' to the other side (started on the left, went to the right).
Click to expand...

Look more closely, or look frame by frame in your editor. The sides are labelled. It's a split/screen example (half of it is processed, half of it is original) , then halfway through, it repeats with the sides flipped
 
I think its all about the h264 settings Panasonic implemented. If you look at the h264 feature list there is so much to screw up for what could be the cause for the banding. The perceptual based compression methods kick in on dark colors. They think you can't tell the difference between those dark blues, so they reduce the information in this area and that leads to banding.

edit: the h264 specs would allow 4:4:4 color resolution: " …
High 4:4:4 Predictive Profile (Hi444PP)This profile builds on top of the High 4:2:2 Profile, supporting up to 4:4:4 chroma sampling, up to 14 bits per sample, and additionally supporting efficient lossless region coding and the coding of each picture as three separate color planes. …"
I doubt that this would be possible within the AVCHD specs with its 24mbit limit. A firmware fix could change a lot tho ...
 
Last edited:
aemonk said:
I think its all about the h264 settings Panasonic implemented. If you look at the h264 feature list there is so much to screw up for what could be the cause for the banding. The perceptual based compression methods kick in on dark colors. They think you can't tell the difference between those dark blues, so they reduce the information in this area and that leads to banding.
Click to expand...

If you're referring to the deep color banding , it's not from the AVCHD compression

The HDMI samples recorded losslessly or with prores also have a form of this banding. This indicates some problem upstream, like sensor or processing issue. Other cameras don't have it (or rather, don't have it that bad)

The other form of "banding" can partially be from the AVCHD compression. You can see this in other cameras, like AF100. But the HDMI recorded from that does improve that type of banding (the "regular" kind , from compression)
 
someone was telling, that the gh2 pulls a lot of resolution out of the green channel, what led to those "I like the canon colors better" discussions. Could be that the internal compensation, the stretching of blue channel, causes this specific banding issue ...

here is the info about the green channel thing:


  • Panasonic has developed a 4K x 2K image sensor (approx. 12 megapixel) based on 'dual exposure technique'
  • The sensor is described as having new architecture that tries to improve sensitivity by a factor of 4 or more
  • The architecture change is a new design which allows green pixels to be read independently of the blue and red ones on the sensor.
  • The green pixels are exposed 4 times longer than the red and blue, thus achieving the sensitivity gain
  • Motion artifacts are introduced as a result, but clever signal processing of the red and blue channels corrects the motion artifacts introduced in the green channel by the long exposure
  • Red and blue channel gain achieved by pixel binning in 4x4 mode

 
Last edited:
aemonk said:
someone was telling, that the gh2 pulls a lot of resolution out of the green channel, what led to those "I like the canon colors better" discussions. Could be that the internal compensation, the stretching of blue channel, causes this specific banding issue ...

here is the info about the green channel thing:


  • Panasonic has developed a 4K x 2K image sensor (approx. 12 megapixel) based on 'dual exposure technique'
  • The sensor is described as having new architecture that tries to improve sensitivity by a factor of 4 or more
  • The architecture change is a new design which allows green pixels to be read independently of the blue and red ones on the sensor.
  • The green pixels are exposed 4 times longer than the red and blue, thus achieving the sensitivity gain
  • Motion artifacts are introduced as a result, but clever signal processing of the red and blue channels corrects the motion artifacts introduced in the green channel by the long exposure
  • Red and blue channel gain achieved by pixel binning in 4x4 mode

Click to expand...

Interesting, thanks for the info

If this is contributing , then a simple test would be a static tripod shot, pointed at the sky. If the "color banding" still persists, then it's not from the motion component and channel exposure differences. Certainly the motion in the sky trails video enhances the appearance of the bands to the viewer.

uwe's test was largely static, and the normal exposure shots had mainly "normal" banding, treatable if you dithered and encoded it better. The end of the clip (the low light) suggests the severity of color banding is inversely proportional to the amount of light
 
PDR said:
So why is there "color banding" in some GH2 videos, but others, like wanderer63's , do not have the superfluous "color bands" ?
Click to expand...

One possible reason is that wanderer63's GH2 video was made with a GH13. :happy:
 
Well I'm certainly not a scientist, but to my eyes this as no relation to color at all, it's a complete codec breakup that looks like any usual compression for web or whatever. Image starts bleeding in ANY bokeh, any color.

Again maybe I have a worst-case unit, i dunno, but I would take these issues into SERIOUS consideration before buying GH2 as a professional or doing any kind of client work. I will keep doing some more testing during the week, also banding comparison to GH13 (whom clearly benefits from the higher bitrates). Will try getting a video up but I think what examples has been posted in this thread is "proof" enough, it's nothing about compression in jpg or vimeo, this is how it looks straight from the camera aswell.
 
Last edited:
goodgoings said:
Are you talking about the screen grabs he put up in this thread? The files are reduced and I can see some posterization. Maybe it's jpeg compression, but if Gunther says the original is clean then so be it.
Click to expand...

He refers to this vid:

http://vimeo.com/22157896

I cannot see any banding in it, but have a look yourself. I shot this with a legacy Canon FD 50mm f3.5 macro lens, 24pH, smooth film mode, everything -2 except color, full manual, f3.5 - f5.6, 1/40th shutter, 5600K, always slightly underexposed according to the camera meter. One suspicion I have is that the banding is much less with legacy lenses, as I read in a post of Lpowell (can't find it now) that the lens construction (and the way how light is allowed to enter) of legacy glass is entirely different from Panasonic and Olympus µ4/3 glass.

www.vimeo.com/gunthermachu
 
There was a lot of discussion on this topic, when the GH1 was first released. I tried several methods including encoding to 10bit codecs, Neat Video, adding grain and FCPs reduce banding button :-)
The hack made a big difference though. It's still visible, but not nearly as obvious as it was pre hack. Anybody struggling with this would do well to search old posts on colour banding, macro blocking etc.
 
i have a lot of poseterization on skies and blurred backgrounds and slightly underexposed footage with all my Canon FD's
 
Take a look at this video (opening shot of the baby's face). Also at 0:40.

GMC, did you shoot your video with sensor crop? This could be one possible explanation... that the banding would be reduced in crop.
 
Last edited:
GMC said:
Hmm, I don't see it (maybe it's just me). Find some screegrabs from my shoot in India from last week (will post in about a month when shooting is completed), used a Fader ND and a Hoya ND8 filter:
Click to expand...

There is some banding in your shots but not as strong as in mine - its more subtle. I suppose that a Fader ND mitigate this negative effect somewhat. This is an image from the same shooting day as the others - but this time with a fader ND:

without grading


with grading


I think the banding is still there but not as strong as in these test-clips without a fader nd: http://www.vimeo.com/22225257 And perhaps it becomes more visible the more blue gradients are in the shot...
 
Have you guys noticed that a quick and dirty scale of a RAW still in Mac Preview reproduces the banding effect?

So I am not sure if it is the DSLR scaling or compression or 8bit colour which is to blame. Probably all!?

Check the second video on this page for a demonstration of the banding occurring on the GH2 when record is hit. In Live View, weirdly, it's fine. http://www.eoshd.com/content/616
 
Martin Oller said:
Take a look at this video (opening shot of the baby's face). Also at 0:40.

GMC, did you shoot your video with sensor crop? This could be one possible explanation... that the banding would be reduced in crop.
Click to expand...

About 40% of the shots were with the sensor crop (ETC) mode. But I think you are right, ETC seems to be less sensitive to banding, becauese the screen shots of India which I posted are also partly with ETC.
 
Andrew: As they pointed out in the zacuto shootout it was impossible to tweak GH1 to look like the Canon/Nikon cameras. It seems obvious that GH2 has a much harsher processing of the image w. contrast & color channels, but I think the real drama here is that, even if GH1 had a lot of sharpness in all that mud, GH2's sensor is extremely sharp. It just crunches 8-bit into what it actually is, unlike let's say the Canons, whom are softer due to the upscaling, and the color banding just "stays put" in a much more acceptable way.

Returned my camera today :undecided Was gonna do some comparisons but it is kinda pointless, I mean whatever floats your boat, as always. To me, I can't stand the banding as it is right now, but im sure a forced high bitrate would help these issues.

GMC said:
becauese the screen shots of India
Click to expand...

Interesting... it might just be that the processing is softer when it doesn't use the entire sensor (although not visible in 1080p, which after all, is a pretty small rez).
 
Martin Oller said:
Andrew: As they pointed out in the zacuto shootout it was impossible to tweak GH1 to look like the Canon/Nikon cameras. It seems obvious that GH2 has a much harsher processing of the image w. contrast & color channels, but I think the real drama here is that, even if GH1 had a lot of sharpness in all that mud, GH2's sensor is extremely sharp. It just crunches 8-bit into what it actually is, unlike let's say the Canons, whom are softer due to the upscaling, and the color banding just "stays put" in a much more acceptable way.
Click to expand...

I asked this before, since the AF-100 is also 8-bit, shouldn't it also have problems with gradients? I don't hear many people with the AF-100 complaining about it.

I'm glad to see Andrew talking about this on his blog:) Hopefully more people with some media influence can get Panasonic on this problem.

Martin, sorry to hear that you returned your camera. I've been waiting so long to get a GH2, but I want to shoot shorts and documentaries and I'm worried about this problem. What are you thinking of getting next?
 
goodgoings said:
I asked this before, since the AF-100 is also 8-bit, shouldn't it also have problems with gradients? I don't hear many people with the AF-100 complaining about it.

I'm glad to see Andrew talking about this on his blog:) Hopefully more people with some media influence can get Panasonic on this problem.

Martin, sorry to hear that you returned your camera. I've been waiting so long to get a GH2, but I want to shoot shorts and documentaries and I'm worried about this problem. What are you thinking of getting next?
Click to expand...

The AF101 uses 10 bit internal processing and compresses the upper 2 bits into a 8-bit signal (like broadcast cameras do). That signal is fed to both the HDMI/SDI outputs and AVCHD engine. The GH2 on the contrary uses 8 bit image processing all the way, so it can only record gradients of 256 steps (also on the HDMI output).

I have no GH2 solution for this; anti-banding filters all work with the principle of blurring, which i don't see as an option. Panasonic will never solve this, because it's likely to be a hardware issue.

Martin Beek (not to be confused with the other Martin in this thread).
 
goodgoings said:
I asked this before, since the AF-100 is also 8-bit, shouldn't it also have problems with gradients? I don't hear many people with the AF-100 complaining about it.
Click to expand...

Sure it does , and yes there was complaining. There were several threads discussing this and ways to improve it including ideal camera settings , using external recorder , post techniques. It was a "hot topic" for a while.

e.g.
http://www.dvxuser.com/V6/showthread.php?235370-Dialing-in-the-AF100
http://www.dvxuser.com/V6/showthrea...ot-with-Nano&p=2239992&viewfull=1#post2239992
http://www.dvxuser.com/V6/showthrea...ot-with-Nano&p=2240154&viewfull=1#post2240154
 
You must log in or register to reply here.
Back
Top