Andrew Reid just posted about a possible way to improve visible banding in 8-bit footage, by adding a small amount of very fine grain. Obviously, this isn't actually bringing back any detail that was lost because of 8-bit, but visually it seems to work.
What do you guys think -- is he on to something? Is "dithering" an 8-bit image by adding noise really a feasible way to fix banding? The results he posted looked good, but the idea of improving an image by adding noise is not something that sits well with me.
Source: http://www.eoshd.com/content/12923/cure-banding-dslr-footage-gh4-4k-holds-key
If you just replace the entire image with a single solid color, you will no longer have any problems with banding, chromatic aberration, softness, coma, flare, or any other number of maladies.
So yeah, you can sometimes fix one problem by adding in a completely different problem.
@sangye yes, adding noise will "fix" banding. Dithering is a common process and has been around for some time. If you've ever tried to create an 8-bit gradient ramp in After Effects, you'll often get terrible banding. You can add noise but the ramp effect has a dithering option built in.
As for fixing banding in video, the results should be similar but I can imagine some shots where the added noise would not be ideal. However, adding 3% noise should be a nice balance between too much noise and fixing the banding issue.
+1 Xenocide38 I've been using this technique since a long time and it works perfectly in order to eliminate banding present in smooth gradients, even when they are rendered directly from After Effects. Mostly with motion graphics. It has been taught to me by an expert photographer who (despite using an Hasselblad camera) uses it to deliver jpeg and tiffs. It is nothing new, nor it is bullshit
I've worked in Motion Graphics (as well as film) for years and its a tried and tested technique of masking 'banding' in 8 bit files to add a tiny bit of noise / grain to any gradients.
I'm also thinking in patenting a twisted coil inside a vacuum crystal ball that when offering resistance to electric current gets "HOT" (technical term meaning Hysteric Omelet Transfusion) ... and emits a very pleasant light. I'm thinking in calling it tungsjuana bomb; the thing is, other than subscribing to my website of course, you'll need an arri alexa and 3 virgins to be able to make it work. Did I mention it cures prostate cancer just by using it 5 minutes a day?
Here's a prototype I'm working on
2012
It's so childish how people look for opportunities to kick dirt on Reid no matter what. If the tip helps you GREAT. It won't cost you anything to try it and see. Often Reid is just passing along tips he's picked up. I take a dim view to those who bash a man that puts in a ton of work and actually contributes to the community at large. No one is perfect but we all have something we can add to the community. I know i've actually benefited from many things he's written about over the years. Just a little mutual respect is all that's required. Rant over!
applying fine noise to fix 8bit banding is nothing new. It doesn't really help if the banding is backed into 8bit VDSLR footage. The added noise can make things even worse if you have to deliver a highly compressed h264 version of your dithered clip, because the added noise can cause sever macro blocking. And depending on the way grain/noise is added, you are getting long render times too. If you get banding from a computer generated element like a gradient or a feathered mask, it helps a lot to render in 16bit even if you are rendering into a 8bit codec/file. Improving the look of CGI elements by adding some grain is a must though.
Dithering aka. adding a little noise can fix banding issues, but depending on your delivery channel noise can be a problem. If you delivering low bandwidth MPEG4/ h.264, or your content will be reconverted (like youtube, digital sat/cable) any kind of noise will be interpreted as details. Due to this fact most compressors do their own denoising before compression. Mostly kinda rude style denoising.. So, if you are aiming for a clean look and want to keep noise levels as low as possible I suggest to apply dithering onto a second layer with keyed gradient/problem color range. Using tools like MB Denoiser in the same fashion can also help alot, just don't forget to add at least the amount of noise the rest of your picture contains so it does not stick out.
@Aria don't get worned out because of stupid people, we are too many :P
Now, personally - and not that I need to explain myself - I was just having fun at the way things are presented/wrapped, both in this thread and by monsieur Reid, god conserve him healthy (no irony here). If you are talking to me, please do adress me, without innuendos, as it's already difficult not to share same physical space. If not, great!
Not that anyone possibly cares, but relating Reids, Laforets, Blooms and other royal families, for me it's very clear: you trying to sell your stuff/ways and promote yourself, gooodd, but with fame and glory also comes exposure, fans and detractors... it's part of the job. Anyway as I've said before, good or bad, important thing is putting people talk about you/your product, je je.
@Alex true, but nevertheless there seems to be a tolerable threshold with certain amount of grain, good x264 settings and (say) vimeo compression.
@RRRR if you want, for a quasi symbolic price, say 128 kronors, I sell the ™ to you :>
I found this interesting and somewhat related.
http://cineform.blogspot.com.au/2014/02/quicktime-16-bit.html
...I'm just surprised some of you never heard of this before. Don't kill the messenger...as Aria suggested he's just passing along info that he picked up somewhere else.
And I too hate the Andrew Reid bashing. I don't follow any of them (Bloom, Reid, Laforet etc.) but I do appreciate the info they give. This is pointed at no one in particular. I just see it forum after forum.
And I too hate the Andrew Reid bashing. I don't follow any of them (Bloom, Reid, Laforet etc.) but I do appreciate the info they give. This is pointed at no one in particular. I just see it forum after forum.
I have goo idea. Let's stick to the topic instead of Voldemort discussions.
"I suggest to apply dithering onto a second layer with keyed gradient/problem color range." - @Meierhans
Great tip. I didn't realize all of these were things people were doing to mitigate banding. Lots of good info here.
thanks @rambo that was interesting!
I think we've been discussing this on and off for a long time, and yes, sensor noise (which is then noise-reduced in post) can be very effective at getting rid of banding because the noise is adding dither to the signal before it's encoded by the camera and hence disguises the banding. I'm sure I posted up some examples at least two years ago comparing a noisy original with subsequent noise-reduction against the same shot but with less native noise which had more evident banding. Although it was a while ago, and I can't find them at the moment. In practice therefore, I don't worry about noisy images as noise-reduction can help with both noise and banding.
It looks like you're new here. If you want to get involved, click one of these buttons!