What causes banding?

[steve green]

I just saw the final edit of a piece I shot with lots of backlit skies and sun flares. There's banding in almost all of the backlit skies. We shot this on three Epic Dragons, 5K, 5:1 to 9:1(frame rate) with Canon EF lenses. I've shot this same scenario countless times in the past an never had I'd seen this before. I have seen it with Sony A7S footage but never Dragon footage. Just trying to figure this out.....Thoughts?

 

[Jacek Zakowicz]

Banding is usually a bit depth issue

 

[steve green]

Understood, so why am I seeing it in graded Dragon footage?

 

[Hrvoje Simic]

Banding is caused by "thinning" of the digital material through image manipulation, where image data is "stretched" beyond material's ability to display in-between tonalities. Since gradients are made of neighboring tonalities they are most susceptible to banding. It should not be present with Epic footage, unless you are watching some poor compression export. Check the signal chain, post processes used and if some generic LUT was stuffed somewhere.

 

[Bob Gundu]

Banding can be hidden with some grain.

 

[Nick Morrison]

Sounds like an issue in the post chain - in the edit, the grade, or final compression, looks like something happened...way after you shot it.

 

[Sabovic Adis]

Banding is intrinsic to Digital - it's just more or less visible.

 

[Graeme Nattress]

Please send me output image that shows the banding and the source R3D frame (one example should be enough). It's highly unlikely that any banding such as you're descibing is coming from the camera.

Graeme

 

[Marc Wielage]

I just saw the final edit of a piece I shot with lots of backlit skies and sun flares. There's banding in almost all of the backlit skies.

What kind of monitor are you using to judge the material? Is it a 10-bit monitor? How is it color managed, and is it calibrated?

What kind of software is being used to color-correct the final edit? It is possible to force all color-correction software I know of in 8-bit, and you will get banding due to "lack of bits" from that problem. I would look at test patterns through the same signal chain and see if you see the same kind of banding there, particularly from gray-scale ramps and anything that would really show lack of bit-depth due to "stepping."

 

[Craig Parkes]

If it's the final edit the banding has almost certainly be introduced in the color grading or export of the video where it was dropped from 10bit down to 8 bit (presuming you are watching online) If you are watching on say your laptop in a video with a 10 bit codec such as Prores the banding could be in the video or conceivably could be limited to your screen not being able to display 10 bits and the banding is happening in the signal chain between the file and the screen.

Most of the time though banding is introduce in the encoding process, going from a good looking 10 bit master down to h.264 can introduce banding if you don't first dither the skies and people generally treating encoding for web as a set and forget process - the skills and knowledge to ensure good quality encoding for web occurs is not really valued as there is never any formalised QC process between encoding and uploading.

 

[M Most]

What kind of monitor are you using to judge the material? Is it a 10-bit monitor? How is it color managed, and is it calibrated?

What kind of software is being used to color-correct the final edit? It is possible to force all color-correction software I know of in 8-bit, and you will get banding due to "lack of bits" from that problem. I would look at test patterns through the same signal chain and see if you see the same kind of banding there, particularly from gray-scale ramps and anything that would really show lack of bit-depth due to "stepping."

I would also ask whomever did the finishing what kind of debayer they used, and whether they made a mezzanine file that was used for the final finishing (and what that format was).

 

[steve green]

If it's the final edit the banding has almost certainly be introduced in the color grading or export of the video where it was dropped from 10bit down to 8 bit (presuming you are watching online) If you are watching on say your laptop in a video with a 10 bit codec such as Prores the banding could be in the video or conceivably could be limited to your screen not being able to display 10 bits and the banding is happening in the signal chain between the file and the screen.

Most of the time though banding is introduce in the encoding process, going from a good looking 10 bit master down to h.264 can introduce banding if you don't first dither the skies and people generally treating encoding for web as a set and forget process - the skills and knowledge to ensure good quality encoding for web occurs is not really valued as there is never any formalised QC process between encoding and uploading.

After reading all the possible reasons for what I'm seeing, I'm thinking the banding is in fact a product of web viewing of encoding that wasn't done to optimize the footage. Here's the file specs of what I downloaded to view it on my MBP. Could this file size and web viewing be the issue? Do you think this is just a quick viewing upload?

 

[M Most]

After reading all the possible reasons for what I'm seeing, I'm thinking the banding is in fact a product of web viewing of encoding that wasn't done to optimize the footage. Here's the file specs of what I downloaded to view it on my MBP. Could this file size and web viewing be the issue? Do you think this is just a quick viewing upload?

One should never judge mastering quality by viewing a 720p, H.264 Quicktime over the Internet. If you're concerned about mastering issues, ask to see the master in a sensible format for making that kind of judgement.

 

[paulherrin]

i would say that a good quality 1.5 min h264 should be way more than 50mb and i'd recommend better quality than 720p as well, even for web (youtube supports up to 4k)

 

[Derek Carlson]

I believe Youtube limits video to 8 bit which will introduce banding.

 

[Blair S. Paulsen]

Perversely, original footage with more code values/bit depth can actually promote more banding in bit starved encodes. If your final deliverable has to be that compressed, then, as others have suggested, you might get better results by dithering gradients prior to final compression. Would also suggest multi-pass encoding with some manual tweaking and, if at all possible, a higher bit rate. YMMV.

Cheers - #19

 

[Marc Wielage]

I believe Youtube limits video to 8 bit which will introduce banding.

Note that Vimeo Pro does allow higher bitrates, but that's a paid service. And it's still 8-bit, as are almost all streaming services and pretty much all mass-market computer and laptop displays. Heck, HD Blu-ray disks are still 8-bit as well.

 

[Jacek Zakowicz]

Note that Vimeo Pro does allow higher bitrates, but that's a paid service. And it's still 8-bit, as are almost all streaming services and pretty much all mass-market computer and laptop displays. Heck, HD Blu-ray disks are still 8-bit as well.

8 bit is fine- there is plenty of 8bit footage with no banding.
6 bit is usually where it shows.... and 6bit interpolated- the infamous fake 8bit...

 

[steve green]

One should never judge mastering quality by viewing a 720p, H.264 Quicktime over the Internet. If you're concerned about mastering issues, ask to see the master in a sensible format for making that kind of judgement.

In the process of getting that done right now.

 

[Craig Parkes]

8 bit is fine- there is plenty of 8bit footage with no banding.
6 bit is usually where it shows.... and 6bit interpolated- the infamous fake 8bit...

Particularly smooth sky gradients can and do introduce banding when encoded in 8bit, which is then made worse when footage is dropped to a 4:2:0 colour space and macroblocking is introduced by h.264 encoding.

8 bit is plenty of bit depth for most normal display scenarios, but there is the odd exception where the master footage breaks the encoding chain and a solution such as dithering needs to be introduced to help things along.