16:8:8 Color Sampling?

[Michael Olsen]

I recently read a column that contained some information relating to a 16:8:8 sampling codec of some kind. Actually, it referenced a 16:8:8 RAW codec which I understand to be a bit...odd, since measuring RAW in the YUV/YCbCr space is a bit like measuring liquids by the pound instead of by the liter. But I digress...

What's the possibility of this 16:8:8 codec? My admittedly poor understanding of the YUV/YCbCr space has it theoretically capping out a 4:4:4 where there is uniform 1:1 representation of Luma and Chroma - and that it just doesn't get any better.

So I'm kind of wondering...what's the difference? And how is something like 16:8:8 even possible? Four luma signals per pixel? Is it just representative of subsampling?

 

[Bruce Allen]

http://provideocoalition.com/index.php/sony/story/wherefore_444/
...explains their viewpoint:

"We proposed that if “4” served SD, then the 1080-line HD world should reflect the far higher sampling frequency of 74.25 MHz. The arithmetic is 74.25/13.5 x 4 = 22. Based on this impeccable logic, we told anyone who would listen that the 1080-line HD counterpart of 4:2:2 should be called “22:11:11.

Nobody listened.

The result is that the nomenclature has lost any tie to sampling frequency and is applied to any image, regardless of resolution. In today’s crazy, mixed-up world, 4:2:2 describes all three of the following images:"

"The ratios ignore differences in image sizes. In the days of standard definition, the “4” specifically referred to the ITU-R BT.601 sampling rate of 13.5 MHz. In 525-line 60-Hz parts of the world, this meant a digital image of 720x480. But today, we’ve got a vast range of image sizes from 1280x720 to 1920x1080, 2K, 4K and beyond. Even as 13.5 MHz passes into memory, the 4:4:4 nomenclature lives on, applied to images of all sizes. This often leads to misunderstandings.

Years ago, Internet chat rooms were abuzz with the relative merits of Sony® HDCAM® 24p recording (sometimes described as 3:1:1) versus Panasonic® Varicam® DVCPRO HD® 24p tape recording (sometimes described as 4:2:2). At the time, some commenters supposed that the “3:1:1” format must have fewer color samples than the “4:2:2” format. Because of different image sizes, it wasn’t and isn’t true."

I merely quote it, I don't necessarily agree with it.

Sony is still bitter that its 1080p HDCAM was dissed for being "worse color" 3:1:1 against Panasonic's "better color" 4:2:2 720p. When the 3:1:1 1080p signal actually contained more Chroma info.

And they're bitter that HD was compared to SD.

This time around they have a "8K" camera which records a 4K signal at 16bits per channel, with Y every pixel and chroma every second pixel.

So presumably they decided it was time for bigger numbers. 16 because... it's 4 times more resolution overall? because it's 16 bit?

No idea.

It's just a ratio, I guess

But they're trying to make things very confusing.

Bruce Allen
www.boacinema.com

 

[Stephen Gentle]

Weren't these ratios initially designed to denote chroma subsampling in YCbCr colour spaces? I see no reason that a RAW camera should have anything to do with them.

All I see is Sony trying to make it sound their camera is better because it has a bigger number (which is stupid in a ratio because mathematically 16 to 8 to 8 is exactly equivalent to 4 by 2 by 2)...

 

[Michael Olsen]

Weren't these ratios initially designed to denote chroma subsampling in YCbCr colour spaces? I see no reason that a RAW camera should have anything to do with them.

That's my understanding...RAW is just RAW. After its processed, it ends up in a codec which then can have some YCbCr color space applied...in which case, 16:8:8 certainly sounds better than even 4:4:4*. Of course, just because something is bigger it doesn't mean better, which seems to be argued in that PVC article.

What I can glean from the article about the change in nomenclature is that the N:N:N system of measurement really doesn't make any true sense as soon as any variance in image is presented. In other words, it's impossible to say one camera that is 4:2:2 would be better or worse than some other camera which might be 8:4:4 or 56:10:8 or any other odd arrangements. Just as well...I'm looking forward to RAW.

*Given what I've been reading, taking 4K 16-bit RAW footage through ProRes444 would actually net a comprable 16:16:16 file because it is 4K? Or wait...that would be 22:22:22 for 1080p. So 4K would be much higher dependent upon the N MHz sampling rate of 4K footage.

 

[Alexander Ibrahim]

Its just evidence that we need new terms and numbers to represent the true nature of digital technology.

OK, so bit depth in bits per channel, effective resolution and percentage chroma data ?

The a 1080p camera recording ProRes 444 (Alexa) would then be 10bpc 1080p 100% chroma.

A camera like the F3 with a ProRes recorder like the KiPro Mini would be 10bpc 1080p 50% chroma

An XDCAM EX camera, like the F3 with internal recording, would be 8bpc 1080p 25% chroma

Something like that.

 

[Gavin Greenwalt]

Video is dead. May chroma sub sampling formats follow shortly thereafter.

 

[George Butterfield]

Or you can always just, "Watch the movie".

 

[Graeme Nattress]

So Sony have invented some new terminology for chroma sub-sampling notation, which is an area of great confusion to begin with.

Bruce: "Sony is still bitter that its 1080p HDCAM was dissed for being "worse color" 3:1:1 against Panasonic's "better color" 4:2:2 720p. When the 3:1:1 1080p signal actually contained more Chroma info." DVCproHD 720p was 960X720 on luma, 480x720 on chroma for it's 4:2:2. HDCAM was 1440x1080 on luma, and 480x1080 on chroma. Same horizontal chroma resolution.

Now Sony are indeed up to their usual games with their current announcement. Some things in this industry don't ever seem to change.

Graeme

 

[paulherrin]

i've made my thoughts about sony pretty clear already. can you guys help me in my frustrations by not posting every piece of garbage they throw out?

please.

 

[Michael Hastings]

...

All I see is Sony trying to make it sound their camera is better because it has a bigger number (which is stupid in a ratio because mathematically 16 to 8 to 8 is exactly equivalent to 4 by 2 by 2)...

so instead of giving you 16 oz of gold 8 oz of silver and 8 oz of copper for your new epic, I can give you 4 oz of gold 2 of silver and 2 of copper? :smile5:

i've made my thoughts about sony pretty clear already. can you guys help me in my frustrations by not posting every piece of garbage they throw out?

Sony has been a major innovator and major force in this industry for a long time and what they do has a significant impact for both Redusers and their clients. It is a competitive business and awareness of the state of the art is important to many of us.

 

[George Butterfield]

so instead of giving you 16 oz of gold 8 oz of silver and 8 oz of copper for your new epic, I can give you 4 oz of gold 2 of silver and 2 of copper? :smile5:

No, nobody mentioned ounces, it's twice as much gold as silver, twice as much gold as copper. Copper and Silver, same amount. A ratio 4:2:2, or 2:1:1, or 16:8:8 or 1024:512:512. You have to watch the movie.

 

[Graeme Nattress]

OB1 waves his hands: These are not the chroma sub-sampling ratios you're looking for.

 

[Michael Olsen]

Thanks, everyone, for your replies. I think I'm prepared to remain more or less ignorant about all of this luma/chroma subsampling since it seems to be a largely fabricated and manufacturer-based theoretical experiment at this point. Best, I think, to just shoot my own tests and see what I see.

 

[Graeme Nattress]

As always measuring yourself is the way to go.

Graeme

 

[Jeff Kilgroe]

I think everyone is reading way much into this. It's nothing more than Sony marketing BS, same stuff they always try and pull over people. They're still living in a 1080p world, still thinking on 1080p terms. Now with a camera that works with a 4K/quad-HD, that's 4X the amount of information, therefore everything else needs to be multiplied by 4, right??? Seems like a logical thing for a marketing executive to assume, right? So now, when you multiply your output resolution by 4, you must (in their minds) also claim your 4:2:2 is really 16:8:8. While they're at it, they may as well claim they're also the first to offer 4K acquisition... Oh, wait, they are.

16:16:16 chroma sub-sampling will be available via a firmware upgrade. And for the unbelievably low price of $9995. Just watch...

 

[Tom Lowe]

Sony continues the nonsense...

 

[Tom.Wong]

I'm just imagining a whole pack of very knowledgable people at NAB surrounding the Sony rep, and putting him or her in the biggest hot seat ever. I really don't envy that person's job right now to have to explain 16:8:8 to professionals who know the ins and outs of all this stuff. I wish I could be there!

 

[Alexander Ibrahim]

so instead of giving you 16 oz of gold 8 oz of silver and 8 oz of copper for your new epic, I can give you 4 oz of gold 2 of silver and 2 of copper? :smile5:

It's hard to believe you aren't goofing. Of course... math education sucks so badly in the US, I say as an American, you may not be.

It's a ratio, not a scalar.

16:8:8 = 8:4:4 = 4:2:2 = 2:1:1

Just means for every two of the first quantity there are 1 of each of the second and third quantity.

So, if you were giving me 16oz of gold for every 8 oz of silver and 8oz of copper I give you, then I'd be perfectly content for you to lower our trade rate to 4oz of gold for 2oz of silver and 2oz of copper

This particular terminology has been stupid ever since digital came to pass. It survives because the idea behind it is important- even if it manages to be mangled by the market.

The only important thing to think about is how much relative color data there is to total image resolution.

This doesn't matter too much when you are LOOKING at the pictures. This is because the people who came up with chroma subsampling actually understood human perception enough to see that we have much less sensitivity to some portions of the color spectrum.

That's why so many people are completely thrilled with XDCAM EX, which is 35Mbps 8 bit per channel 4:2:0 ... which is a a system that collects 25% color data resolution.

Now, the computers we use to process images "care" a LOT more than we do.

It's because of how they do the math - which is more related to the reality of the scene than our perception- so all the perceptual tricks that look good to us low speed meat computers crumble pretty quickly.

So, when you are pulling keys in fine detail, or making large color adjustments or any number of other post processes the less "real" your image data is the more quickly it falls apart.

You've seen the results- banding/posterization and blocky key edges amongst others.

So that's it, that's why the RATIO, not the gross scalar amount of color data matters.

That's also why Sony's 16:8:8 is market speak. Its mathematically no different than 4:2:2 . or even 2:1:1.

The upshot, and what's interesting about the Sony prototype is that they are starting with 8k and aiming for a 4K 4:4:4 deliverable.

They are actually collecting data for every point of the eventual target resolution, and oversampling the resolution significantly.

In some ways, they are trying to out-Red Red. Don't confuse deep pockets and following the competition because of fear with innovation though.

I agree that what Sony does has an impact on all video/film users and their clients.

I have to disagree with the that Sony is an innovator. Sony has focused on strangling innovation in order to keep prices high whenever possible.

Really, if Red is Apple, then Sony is Microsoft. Copying away while trailing and claiming that they invented everything they implement no matter how blatant the copy is.

Remember that this is the company that, after the release of the Red One, still advertised the EX1 as "too much camera for the money."

It may sound like a joke, but they were not kidding.

In fact the EX series (and its descendant the F3) are some of the most innovative cameras Sony has delivered in a very long time. Ever since the first "Star Wars Cam" in fact.

So to sum up in the last decade Sony has gone from HDCAM to HDCAM SR and from 8 bit 1440x1080 "3:1:1" to 12 bit 1080p 4:4:4.

Yay?

Let's have a look at Red. Red One delivered 12bpc 3.2k with more than 75% chroma in 2007. Epic is (as I understand it) 16bpc 4K+ with 100% chroma in 2011, with full time 60fps HDR available in a camera half the size.

Red is the industry innovator, and technology leader. By a LONG shot. They are the only ones presently following the curves of the fundamental technologies. IT tech, not old industrial tech.

Sony isn't stupid, just arrogant. They are finally getting in gear and doing something interesting. Good for them and the industry.

I don't think it matters though. Their camera probably won't be a perceptual improvement over the Epic, and will probably cost a lot more.

I think for most viewing environments, including most "real" IMAX theaters, 4K delivery is a "retina display", i.e. one where your eyes can no longer perceive improved resolution.

So ... I hope that answers your questions. (I'm talking to the OP here, Michael Olsen)

 

[Michael Olsen]

Hi Alexander, thanks for your reply. I'm beginning to think that, as I move to a RAW world, this YCbCr and other "video" issues is increasingly becoming a non-issue.

With RED, it seems the camera just gives you all there is with a user-selectable degree of compression. Ironically, it's much simpler than the alternative.

--

Also, I really didn't intend on this thread becoming a big thread about Sony. I was really more interested in the terminology. Unfortunately, I now understand that this terminology isn't something that is set by some third-party committee, but is evidently up-for-grabs to be used and altered by manufacturers as they see fit.

 

[David Mullen ASC]

Yes and no, you still have to deliver versions of the project in video, and even if chroma subsampling notions are becoming less relevant in terms of capture when shooting RAW, it's useful to always keep in mind that to create a color image you need three channels of info -- red, green, and blue -- and they will have some differences in resolution, etc.

Now that's true with film technology as well, not all color layers resolve detail to the same degree, but it is still useful to keep in mind that single-sensor camera systems need to do something to generate RGB from some sort of color filter array over a monochrome sensor.

As for Sony thinking in video broadcast terms, I would hardly expect otherwise, that's their background. Doesn't really matter in the long term, the real question is how do the images look? The second one is, how much does it cost? I suspect this camera is going to sell for 3X what the Epic sells for and thus will operate in different spheres of production for most people, especially owner-operator types. I also doubt it will be the same small size as the Epic and thus be less friendly for 3D rigs.