PDA

View Full Version : Banding in GH1 Videos



Daniel L.
04-28-2009, 07:23 PM
I have noticed a lot of artifacts related to banding in every video with movement. You can see it here when the camera pans: http://vimeo.com/4385147

It's possible this is due to the web compression but I have not seen it on web videos shot by other cameras. Maybe it's a problem in some post process.

Would really appreciate it if someone with the camera could comment on this issue. It looks like possibly a big problem.

Kholi
04-28-2009, 07:28 PM
Banding seems to be a direct result of image settings, noise reduction in general.

I've got a lot less banding in my images. It's still present at times, but not as bad.

Daniel L.
04-28-2009, 09:12 PM
Thanks Kholi.

Since the GH1 uses a variable bit rate codec the data rate increases during pans, especially fast pans. Sounds like there is a bottleneck. I suspect the camera is unable to handle the signal. The good news is that they should be able to fix this with a firmware update. The bad news is that they would have to reduce the signal rate.

DavidNJ
04-29-2009, 05:21 AM
It probably isn't a camera limitation. The GOP can provide more detail if there is little movement between frames and less if there is a lot. The firmware fix would be either increase the bandwidth, which may be possible with a clock divider change, and/or offer a native 24p mode, which is probably unlikely.

If there is a way to trick the camera into a PAL mode, there should be a 1080p25@17.4Mb/s, effectively 25% more bandwidth.

Oedipax
04-29-2009, 06:24 AM
I'm sick of this camera already! :-Laugh(DBG)-1

When's the friggin' Scarlet release again?

LizaWitz
04-29-2009, 08:08 AM
If there is a way to trick the camera into a PAL mode, there should be a 1080p25@17.4Mb/s, effectively 25% more bandwidth.

Nonsense.

Barry_Green
04-29-2009, 09:26 AM
If there is a way to trick the camera into a PAL mode, there should be a 1080p25@17.4Mb/s, effectively 25% more bandwidth.
Doesn't work that way, and here again is a prime example of why talking about numbers is nearly completely meaningless. You're talking about AVC, one of the most advanced codecs on the planet. Field duplication is probably the easiest thing in the world to implement, and I'd be surprised if the codec is wasting 1% of its bandwidth on the duplicated fields.

Again, not to scream it too loudly, but: TALKING NUMBERS LIKE THIS IS POINTLESS. I mean, you could just as easily say that 720/60p will be 2.5x as compressed as 720/24p. It isn't. Why? Because the additional frames, 2.5x as many frames, are very mildly changed between each other, and therefore compress astonishingly efficiently! Whereas the amount of temporal difference between 24p frames makes the codec work harder. If there's a degradation of codec efficiency in 60p vs. 24p, I'd guess that it's not more than about 10%, versus the 150% you'd think by just looking at the numbers.

So would 1080/25p be more efficiently compressed than 1080/24p @ 60i? Yes, probably. But 25%? Not on your life. Maybe 1% or 2%, but that's probably about it.

DavidNJ
04-29-2009, 09:46 AM
You're right...I didn't account for the long GOP...but did look at the bit rates reported earlier by others. At 17Mb/s, it would be equivalent to 2:3 pulldown or 30p at around 22Mb/s.

djkarn105
04-29-2009, 11:46 AM
This thread has got me thinking a bit.

In the end what I care about is the final image and, even if this camera does have full manual controls, if the image has a lot of artifacting, is super crunchy and can't be decently graded in post, I think I'd rather deal with the workaround of the 5d Mark ii if its image hold up better.

Daniel L.
04-29-2009, 12:00 PM
It's still early, people are just beginning to learn how to use the camera. All we have is numbers and there is more to it than factory specifications. We really have not even seen any "raw" samples.

It's not yet clear which camera performs better. I did not mean this thread to start speculation, lets be patient.

Michael Olsen
04-29-2009, 12:02 PM
This thread has got me thinking a bit.

In the end what I care about is the final image and, even if this camera does have full manual controls, if the image has a lot of artifacting, is super crunchy and can't be decently graded in post, I think I'd rather deal with the workaround of the 5d Mark ii if its image hold up better.

The vast majority of videos posted up, along with initial testimony from more experienced users, makes me think that, for many uses the GH1 will output a video that meets or exceeds necessary standards. As Kholi stated in his thread:


3. Rolling Shutter = RED ONE spec...

4. Image Quality = Sucker looks like EX-1/RED @ 1080. Straight out of the HDMI into a Production Monitor all I gotta say is "fin~"

5. Manual Control Extent = EVERYTHING. I do mean EVERYTHING.

6. Custom Color Profile = ...YES. And just wait til you see this stuff with the NOISE REDUCTION off. Yeah, it comes standard ON with this camera. >=D

7. Shutter speeds: 1/40, 1/50, 1/60 on Up. Close enough to 1/48 for me. Although there doesn't seem to be a way to avoid those pesky Sodium Vapor lights or Fluorescents yet. Gonna have to dig further.

8. ISO - 100, 160, 200, 250, 320, 400, 500, 600, 640, 800, 1000, 1250, 1600 in Video Mode

9. Flip Out LCD - ...The LCD is awesome.

10. Overheating -- Camera's been on for an hour so far. No overheating. Awesome feature on this: it turns the LCD off until the lens picks up motion, then it snaps back on. Sweet.

11. KIT LENS -- F4 indeed. And I have no reservations with using it for as much as I can. If this is the limitation of an F4 then it's all up hill from here man. The Lens isn't really friendly for manual use, unless you wear it in. But this sucker is QUIET as hell.

...

13. COLOR PROFILES -- You can save custom versions of Panasonic's curves. There are SEVERAL, and versions of Black and whites as well.

Assuming the camera continues to hold up to these initial impressions, it should be very suitable for many things. It has it drawbacks though, and they will begin to surface as people put it through its paces properly. High motion shots, as Kholi tested in his car, evidently tear the codec up rather badly, making them unusable. The banding we are seeing could be part of the codec and a real limitation, or it could just be a factor of a user picking the camera up and shooting with it right out of the box, not bothering to set "ideal" white balance, ISO, etc. Fluorescent flicker, without many more precise shutter speeds, may also prove to be a long-term pain. Or, these problems could all be created or worsened by the use of vimeo and YouTube compression.

Really, we need to see the footage the three DPs are working on (and hear their commentary about actually shooting and posting it) before making hard judgments - most everything is still guesswork. If I had to bet right now, I'd say that the MkII will be better than the GH1 at some things, but that the GH1 will be better at others. Like anything, the camera you choose should be the one that offers you the most benefit. Just tools in a toolbox.

Eddy Robinson
04-29-2009, 12:13 PM
(...)

:-Mark-13(DBG) Look, there is not a linear relationship between the bandwidth and the number of frames in the 50i/60i wrapper. The wrapper is just indexing scheme that defines how the datastream will be parsed by the receiving hdmi device. this is just as fallcious as your conclusion that the data rate was going to be 13.7mbps based on a screenshot of someone's ProRes workflow.

LizaWitz
04-29-2009, 12:19 PM
It's possible this is due to the web compression but I have not seen it on web videos shot by other cameras.

This kind of artifact is present in every YouTube and just about every Vimeo video I've seen online. It is the low rez compressor for web streaming video that's causing it, possibly due to the way flash decodes it or due to the simple fact that the video has been re-encoded for low bandwidth.

Park Edwards
04-29-2009, 12:22 PM
i love the assumptions based off of youtube and vimeo stuff....really? one raw clip has surfaced...ONE...true .mts video and there was no banding. get over yourselves already.

Daniel L.
04-29-2009, 12:22 PM
This kind of artifact is present in every YouTube and just about every Vimeo video I've seen online. It is the low rez compressor for web streaming video that's causing it, possibly due to the way flash decodes it or due to the simple fact that the video has been re-encoded for low bandwidth.

That's what I suspected at first but Kholi did confirm that there is banding present in the original videos.

Daniel L.
04-29-2009, 12:24 PM
i love the assumptions based off of youtube and vimeo stuff....really? one raw clip has surfaced...ONE...true .mts video and there was no banding. get over yourselves already.

Trying to get this information from reliable sources who own the camera. Not youtube videos. Vimeo link is only there to demonstrate what I was talking about.