Pages: [1] 2   Go Down

Author Topic: Banding-posterization issues  (Read 13160 times)

fredjeang

  • Guest
Banding-posterization issues
« on: February 11, 2012, 03:34:19 am »

Hi,

I'm very concern about this topic.

Until now I've been focussing my learning, for circunstancial reasons, on the post production area, mainly editing footage from others, P2, Sony and Red format. In what filming is concerned, the experiences I had were always in controled light configuration and indoor.
Since I started to involve myself in the shooting itself in more diversified situations, issues appeared more clearly under certain circunstances and the one I'm really concern about, specially after the GH2 bad experience, is the banding-posterization problem. It's hugly and nor the bitrate neither post cure it at all.

The informations I could collect in internet and talking to people on set are showing that the problem comes from a sum of many factors. 8 bits, 4:2:0, codec used, light incidence, plane surface, log etc...

Rainer's footage shows that the GH2 is worse than the 5D2 on that aspect, but the 5D2 isn't free of it either. It just resolves it better but far from being ideal.
But my concern is that recording to an external device in 4:2:2 from those still cameras does yes improve a lot, but still is present to some extend.
I've noticed that the light angle is critical too.

Then, I watched many high-end video advertising campaigns and there is absolutly no banding involved.

Are those dslrs-evils absolutly unable to record a banding free-footage?

----parenthesis-----

I'd like, if possible, Michael to report on the Nex 7 with and without the HDMI recorder on that aspect too because it's an issue that has been completly ignored by reviewers on the GH2 and the only voices you could hear ww were that this camera is a little motion bomb, and then, surprises. It could lead to confusion on what real uses should we expect. When you're experienced, you know those things, but when you're rather new, it's more difficult to get a clear picture. There are so many new things to learn that the attention is focussed on this or that aspect and it's a constant discovery of problems that seems to never end.

Like many people, I've been attracted by the dslr-motion wave and the convergence idea, mainly for the form factor and wanted to see in this fload a suitable solution that technology now allows us to enjoy. But it seems that it's not the case yet. The footage deliver by those mini cameras isn't suitable for many high-end or professional requierements.
As many people who are just coming into motion, I'm learning the hard way, from mistakes, tries and conclusions from field experience. You can be lucky and see an issue immediatly or not, it depends.

But if reviewers all over the world were a little more explicit on certain aspects, it would be really helpull not to have to go through those testings and then realise later that the camera-system involved is useless or really weak. You can see in internet that many GH2 owners have bought this camera because of its price-size-reputation, and then started to discover serious issues later, not without surprise. Those complains are hidden in the middle of the overall crazyness and enthousiasm ww the camera still enjoy to date.

It happened to me too exactly the same way, I've been recommended this camera, I've seen experienced pros using it on set and then the worse is that I also recommended with confidence this camera to some people who asked me for a light reduced high performance motion gear and then realised quickly that it was a mistake because of those output problems discoveries. Lesson learned, never recommend something that hasn't been tested for a long time even if the rest of the world is saying it's the best, better than the 5D2 etc... (without talking about the discutible HDMI implementation etc...). There are things we can sacrifice, but the image quality no.

Don't get me wrong: I'm not asking the reviewers to think for me and avoid field testings. In my case I started late the field testings and those issues actually are much more pronunced under certain circunstances. I'm just asking the reviewers to report as much as they can, when things are not normal and-or could represent a serious uncurable downside. In the case of GH2 it is exacerbated by a factory engineering issue and Panasonic is aware of it.  

Actually, some motion-video pro and one photographer of this forum warned me on the dslr-evil cameras. I thought that it was a glimpse of orthodoxy from their part and was so enthousiastic on the format that didn't listen, I must admit now that they were right.

-----------

This is not a big deal. The camera was cheap, I didn't invest in m4/3 lenses and I'll simply buy a more solid tool. It's just been an incredible time consuming and false economies.

Back to the tread: about those banding posterization problems in general...

Any thoughts?


« Last Edit: February 11, 2012, 04:35:42 am by fredjeang »
Logged

Hywel

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 294
    • http://www.restrainedelegance.com
Re: Banding-posterization issues
« Reply #1 on: February 11, 2012, 09:42:58 am »

Tests I did with my cameras and some help from Barry Green over at DVX user led me to believe that the primary cause of banding is 8 bit colour resolution.

It is made worse by a lack of noise (either from an inherently quieter sensor, or from noise reduction) because that removes some of the random dither which helps break up a posterised/banded portion of the image. So paradoxically I find the effect much more pronounced on my AF100 than on my old HVX200, which had a much grainier image.

Codec didn't seem to have much effect- banding issues from an AF100 on a static scene lit to deliberately provoke the problem showed the same effect recorded on the internal AVCHD coded and recorded externally to much higher bitrate on a Nanoflash. I'd imagine that smearing out the noise with a very compressed codec wouldn't help, and indeed I have seen the AVCHD footage tends to degenerate into mush much faster when there is a lot of motion going on, but in terms of banding on a static scene it made no difference.

I've also noticed doing timelapse on my Canon 7D that shooting JPEG is MUCH more prone to banding than shooting RAW, which supports my hypothesis that it is bit depth is the main factor here.

For now, the main thing I avoid is smooth gradients of light on plain walls, but it is tricky to do in all cases. If it is really bad, I add grain in post to help break up the bands, but this is far from ideal.

The only real solution is 10 bit or better, which has to present through the whole image and recording chain (recording 10 bit ProRes from the 8 bit SDI output of an AF100 doesn't help, although if you do any grading in post you should do it in a higher bit depth working space).

The Panasonic HPX250 is 10 bit to P2 cards.

REDs and Alexas are at least 10 bit (I think Red One is 12 bits, Alexa would be 10 bit if recording to some families of ProRes but capable of more, Scarlet and Epic are 16 bit, I believe). You'll need the data rate to retain the fine detail in the colours, too, but ProRes should be fine.

The 8 bit limitation of the Canon C300 was one of the reasons I didn't consider one, although I gather from posts by people who have used it that the internal quad HD -> 4:4:4: 1080p path and the very film-like nature of the noise means that banding is not as much of an issue for that camera despite the 8 bit clamp. I'll believe it when I see it pointed at a smooth gradient falloff on a white wall ;-)

  Cheers, Hywel.

Logged

fredjeang

  • Guest
Re: Banding-posterization issues
« Reply #2 on: February 11, 2012, 10:36:46 am »

For now, the main thing I avoid is smooth gradients of light on plain walls, but it is tricky to do in all cases. If it is really bad, I add grain in post to help break up the bands, but this is far from ideal.


The 8 bit limitation of the Canon C300 was one of the reasons I didn't consider one, although I gather from posts by people who have used it that the internal quad HD -> 4:4:4: 1080p path and the very film-like nature of the noise means that banding is not as much of an issue for that camera despite the 8 bit clamp. I'll believe it when I see it pointed at a smooth gradient falloff on a white wall ;-)

  Cheers, Hywel.


Exactly.

Same dilema for me.

I've done some pre-prod tests with a fashion photographer in a location that has a white wall in a long corridor with the 5D recorded externaly and the GH2 with 100mb hack. It's crap. Really crap. And the more the natural light comes with a little angle, the crapper it is. Very difficult to compensate with controled lights although possible, but you have to decide if you want to loose the subtle gradient, outside it's simply impossible. (unless you have 3 trucks of cine lightning and 10 people on crew just for that. Holly shit, I understand more and more all the cine heavy circus. We're not in still land and I'm afraid that the convergence, cheap and reduce is not for tomorrow yet)

Although the 5D2 is better, I don't see it neither for advertising assignements.

If you do your indy movies it does have very little impact because in the end you can compeat in festivals, and those issues aren't really a concern but for advertising and other demanding markets, those cameras are simply unusable, with, and without external recorder because the external recorder limit the damages but doesn't really solve them completly. I imagine that the Nikon D4 will be exactly the same saga. Ok if you report news, do your indies, low-end corporate cheap movies but very little more.

I'd also like to see this Canon C300 on a smooth gradient big white wall too, and with a wide. I sincerely don't beleive it neither.

I'd like to have the absolute confirmation that the Red cameras are free of those issues. Please if Red users could comment on that it would be great. I'll do my own testing before buying anyway renting a camera for one day.



So, what in the market today that doesn't cost 100.000 euros can deliver a high-clean quality imagery in motion, free of banding-posterization and artifacts of all kind?



« Last Edit: February 11, 2012, 10:57:05 am by fredjeang »
Logged

Morgan_Moore

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2356
    • sammorganmoore.com
Re: Banding-posterization issues
« Reply #3 on: February 11, 2012, 11:19:40 am »

I wonder how the client is going to deliver all this quality to their customers ?

Not via the web or compressed cable I guess?

Ultimately the acquisition device only needs to out resolve the final delivery

S
Logged
Sam Morgan Moore Bristol UK

fredjeang

  • Guest
Re: Banding-posterization issues
« Reply #4 on: February 11, 2012, 11:57:12 am »

Well morgan, honestly I see it complicated. Not impossible but complicated.
« Last Edit: February 11, 2012, 05:39:47 pm by fredjeang »
Logged

Morgan_Moore

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2356
    • sammorganmoore.com
Re: Banding-posterization issues
« Reply #5 on: February 11, 2012, 03:07:42 pm »

I don't really see the problem

Shoot FS100 GH2 5dmi2 whatever

Get some bigger gigs, buy a Scarlet, rent an epic

Spread over three years the cost of Scarlet is not that high per month

Of course Ive not bought crap stuff Ive bought decent accessories..

Miller 25lb head
V-Lock battery system

Epic drops straight into my workflow..









Logged
Sam Morgan Moore Bristol UK

Hywel

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 294
    • http://www.restrainedelegance.com
Re: Banding-posterization issues
« Reply #6 on: February 12, 2012, 05:08:41 am »


I'd like to have the absolute confirmation that the Red cameras are free of those issues. Please if Red users could comment on that it would be great. I'll do my own testing before buying anyway renting a camera for one day.


On my one day hire of a RED One, I tested for this, comparing with the AF100, and the RED showed no signs of banding issues in my usual shooting spaces.

Hence I have a Scarlet on order. As Morgan says, spreading out over 3 years the monthly cost is quite bearable :)


Cheers, Hywel.


Logged

fredjeang

  • Guest
Re: Banding-posterization issues
« Reply #7 on: February 12, 2012, 07:43:47 am »

Hywel,

It's reassuring. Thanks for the info.


Morgan,

Nice distagon. Is it the Rollei version? Did you find an adapter for the Sony?
« Last Edit: February 12, 2012, 07:45:38 am by fredjeang »
Logged

Morgan_Moore

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2356
    • sammorganmoore.com
Re: Banding-posterization issues
« Reply #8 on: February 12, 2012, 09:19:53 am »

Very nice lens indeed - it had a nice feel

Don't know too much about it it was on the (Epic) camera

Personally I use nikkor lenses mainly 35-70 3.5 and 18mmF4 - which will also work on red products if needed

S
Logged
Sam Morgan Moore Bristol UK

Graeme Nattress

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 584
    • http://www.nattress.com
Re: Banding-posterization issues
« Reply #9 on: February 12, 2012, 10:06:22 am »

We've never had any reports of banding issues with RED footage. We've had the odd report, but it's always turned out to be something in some post application that's used inappropriate bit depth or it didn't dither the image properly on decimation to a lower bit depth, exacerbated by the low noise of the RED image.

8bit is actually fine with regards to video and banding issues (on displays of normal dynamic range capabilities), but it requires that the image be properly dithered. 8bit is probably not a suitable post-production bit depth unless fantastic care is taken over precision and dithering. In reality, 10bit is minimum.

The problem is actually more complicated than just the bit depth, but mostly related to lossy compression. If there is any slight noise or dither in the image, the first thing any lossy codec does is remove that as it tries to get the image to compress. Motion adaptive codecs tend to be the worst for this, but it happens with practically all lossy codecs. The RED codec doesn't create banding when this happens, and that was necessary for it's application as a codec to record raw and be graded afterwards. Most compression research goes into codecs to be used for delivery applications, not acquisition or especially not raw recording.

If banding is an issue for what you shoot - you either have to go uncompressed (and thus keep the natural noise or dither in the signal) or use a codec that had been designed for the purpose.

Graeme
Logged

fredjeang

  • Guest
Re: Banding-posterization issues
« Reply #10 on: February 12, 2012, 12:47:19 pm »

Thanks Graeme,

I agree.

The post is advanced with dither techniques. This simple idea to do a master in, let's say, a DNxHD 175x like I often read and then apply the dither on the 8 bits outputs doesn't seems to me like a satisfactory answer,
it's indeed more complicated. The problem is that we do the editing with different source material on zillion layers, keyers etc...but boosting to a higher bit on post a 8 bits source material won't change anything if the banding is present on the source file.   

I'm not convinced by using a brutal dithering on import for all, this is a kind of violent process (unselective) and what happen if you got your master in 10 bits but then it's viewed later on 8 bits ? That's a difficult dilema to overcome.

We are here in really advanced editing techniques. The all picture is more complex than just adding noise.

Nota: there's a node in Nuke called: Deep Compositing Nodes that is specially teating this aspect.
« Last Edit: February 12, 2012, 01:13:58 pm by fredjeang »
Logged

Graeme Nattress

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 584
    • http://www.nattress.com
Re: Banding-posterization issues
« Reply #11 on: February 12, 2012, 12:59:02 pm »

No, wildly adding dither is not the solution. But upon bit-depth reduction, that should be performed with appropriate dither, but that's not always the case.

Graeme
Logged

Bern Caughey

  • Full Member
  • ***
  • Offline Offline
  • Posts: 179
    • WONDROS
Re: Banding-posterization issues
« Reply #12 on: February 13, 2012, 11:20:39 am »

"Don't let the perfect be the enemy of the good."
Logged

Bern Caughey

  • Full Member
  • ***
  • Offline Offline
  • Posts: 179
    • WONDROS
Logged

fredjeang

  • Guest
Re: Banding-posterization issues
« Reply #14 on: February 14, 2012, 03:56:26 am »

http://en.wikipedia.org/wiki/Canon_EOS_5D_Mark_II#Independent_film_and_television

But the Canon, IMO, handles better the banding issue than the GH2. Testings between both cameras show it clearly and a venerable member of this forum who also did the same comparaisons reached the same conclusion: the GH2 is specially sensitive to banding, more than the 5D2. The GH2 is better than the Canon in many aspects but has issues difficult if not impossible to cure. A part from the banding due to whatever bits-codec etc... reasons, it also has, you know it, another different banding issue on the 3/4 part of the image in higher isos. This has been vastly commented on-line and Panasonic recognized it but never fixed it. That fact almost disable the confident use of the camera in higher isos.
Then, the HDMI output wasn't very "friendly-user" (on purpose) from the release of the GH2. How about external 12V batteries for power on field? The Canon is now hackable and I think that this is a more reliable camera considering global parameters.
So yes, the 5D2 is in the end a much more usable camera IMO than the GH2 (and the form factor too, that has impact on lenses) and you saw that the 5D2 crowd didn't replace suddenly their Canons for this Panasonic. Too problematic.
Despite performing better than the 5D2 in many aspects, I don't think we're going to see a lot of prods shooted on the GH2 like we see in the Wiki link. For me, the GH2 is a great camera, it was almost there, I wish it were, but too toyish and not really cured from its issues. The best image I can find is that the Panasonic is an "unstable" or "unfinished" camera.

For me, I won't invest not one more € on 8 bit dslr-evil cameras present and future. Not because I think we can't do good stuff with them, simply because I want a gear that you can forget about it, that is built to handle whatever conditions, that it's not an endless compromised to manipulate in post, that is working without having to search D.I.Y solutions over the internet, in short: what Coot was saying: there are false economies.
« Last Edit: February 14, 2012, 04:20:44 am by fredjeang »
Logged

Morgan_Moore

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2356
    • sammorganmoore.com
Re: Banding-posterization issues
« Reply #15 on: February 14, 2012, 05:29:54 am »

Fred

I think you should have a good think about the Sony FS100

-You can run from 12v batts

- big HDMI well placed is the best that you can get that is not SDi

-XLR sound (I also love putting one mic into two channels at two levels - great fro the photographer not so confident in sound)

-no moiree

-super efficient codec, long record times

nikon G or manual, canon EF, FD, contax, PL,

-more than one tripod hole

OK yes it is 8bit, yes it will fail the half lit wall test, yes the highlights can go strange

-NO ND is bad

-price point that is not too painful (compared to the 8bit C300 or F3)

and fine quality 50 or 60p (not on the Scarlet/C300/F3?)

The camera is taken some tome to get the colour looking good and learn to 'expose to the left' (!) but I think I am getting it now

Best

SMM
Logged
Sam Morgan Moore Bristol UK

fredjeang

  • Guest
Re: Banding-posterization issues
« Reply #16 on: February 14, 2012, 06:10:51 am »

Thanks Morgan,

This Sony is highly regarded as a great bang-for-the-buck. I considered it.

But it looks like a teenage girl's skin with all those buttons everywhere. It doesn't give me the sensation that it's straightforward-intuitive to operate.
Less is more is a little bit a sentence I should write on the door of my studio.

Also, the Cooter's posts on that camera made me think and reconsider this option a time ago, specially on the point that they find difficult to give it the look they want
or more exactly that it never ends to look very "video-like".

The No ND is indeed bad in this camera format.

But I'll re-look all the aspects and check again.

Cheers.
 
 
Logged

Morgan_Moore

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2356
    • sammorganmoore.com
Re: Banding-posterization issues
« Reply #17 on: February 14, 2012, 07:08:46 am »

Yes it looks like a toy

No there is no ergonomic love

Actaully I only ever hit a couple of buttons,

S+Q - instantly go to 50p (and clips are conformed on the fly as slo mo)

Last scene - playback what you just shotm, (complete with sound) and press again to exit - (with the EX1 you cannot exit!)

Shutter - press and dial in
Wb - press and dial in
Gain - three available by toggle - more in the menu :(

As for the highlights

-I think cap exposure
-desat in post http://dslr4real.tv/index.php?option=com_zoo&task=item&item_id=97&Itemid=1

Its no wonder cam.. but you really don't want a DSLR after using..

S

Logged
Sam Morgan Moore Bristol UK

fredjeang

  • Guest
Re: Banding-posterization issues
« Reply #18 on: February 14, 2012, 08:18:37 am »

I see it that way, at least for me:

I'm not very hot in the idea to invest again into a camera-system that yes, would feature more professional connections etc..., but still would have 8 bits issues to deal with and this ND stuff is really a big missing feature too. I could live with the messy buttons though. It's not an expensive camera but it's already a significant investment.

I rather stick with the GH2, specially after the latest hack I installed from a Bern's post wich is very good, living with its limitations, accepting the banding factor and keep going learning and progressing with it.
While I save all the money for a Scarlet (banks here don't give anymore credit to date unless at conditions I don't agree). If I go today Sony, it means that I'd push further the purchase of a Scarlet but having again, to some extend, a number of issues like the one you posted that are time consumming.

The thing is that Pana, Sony, Canon, have a pro market to protect and tend to protect it with purpose missing features or issues on the image side. The message is clear: "here you got our camera X model that isn't bad but has a lot of problems...want more? it's 20.000 bucks!"

At 15.000 I got a Scarlet in full working order. End of hassles. Peace of mind (at least it seems to be the case)

I would own the Sony I'd do the same. Sticking with it, getting better with it, and saving for a Red or a F3, something much more robust and hassle-free.

I don't see this new Nikon at all as an alternative. Too low bitrate, 8 bits etc...we fall again in the same.

That's the way I see it.
« Last Edit: February 14, 2012, 08:30:34 am by fredjeang »
Logged

Morgan_Moore

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2356
    • sammorganmoore.com
Re: Banding-posterization issues
« Reply #19 on: February 14, 2012, 09:25:10 am »

I think it depends on your business model

90% of my paid work has involved interviews and often handing the rushes to an external editor

At that point XLR + power + long takes = wroth every penny

For an indy movie with a sound person, whatever - its different

I don't the the GH2 but the 7d when the HDMI lead is wobbled the camera stop rolling - just a killer for interview use

As for the costs - agin depends what you are tooling up for

I have a comprhensive rig with lights, $3k tripod, lavaliere mics, ($1k) nice mics ($.5k), many Vlock batteries ($3k with charger)

Many things that make the setup quite expensive with no camera in consideration - all this stuff is more expensive than a bare Scarlet!

But things that mean I am actually ready to roll for money

It depends on your business

I would have a Fully kitted FS100 before a naked Scarlet, .. hopefully in time I can add a decent camera to that kit..

S



Logged
Sam Morgan Moore Bristol UK
Pages: [1] 2   Go Up