16:8:8 Color Sampling?

[Eki Halkka]

I don't care how you cut it... 16:8:8 falls behind RAW for color flexibility.

If, and that is a big if, that number actually refers to oversampled 4:4:4, and (big) if the camera can save linear RGB files with high enough precision with no other processing than demosaicing and scaling, there shouldn't be any perceivable color flexibility difference, as far as i know.

Or am i wrong?

4K is the cinema standard.

...it will be for sure. As for today, depends on how you define "standard".

 

[Jannard]

If, and that is a big if, that number actually refers to oversampled 4:4:4, and (big) if the camera can save linear RGB files with high enough precision with no other processing than demosaicing and scaling, there shouldn't be any perceivable color flexibility difference, as far as i know.

Or am i wrong?



...it will be for sure. As for today, depends on how you define "standard".

Any RGB codec can't compete with RAW... easy to understand.

4K is the future of all distribution... from YouTube to cinema. Trust me.

Jim

 

[Tim Whitcomb]

Any RGB codec can't compete with RAW... easy to understand.

4K is the future of all distribution... from YouTube to cinema. Trust me.

Jim

We do trust you Jim, but this shit makes my head spin. It all sounds like marketing hype to me, and if I had to compete against you.. I would come up with all kinds of mumbo jumbo to confuse people that "if I have more numbers in my specs...
well then people will think I must be better..". it's so LAME... but like you said,

the important part of the announcement is the validation of 4K delivery... that must make you want to break out the Cubans. :sifone:

 

[Eki Halkka]

Any RGB codec can't compete with RAW... easy to understand.

Well, RAW does have it's advantages, like lower bandwidth and debayering choices, but i actually still don't understand the difference as far as color versatility is concerned. Both should give one the same control, assuming no funky in-camera gamma etc. adjustments being made, and the RGB being bit-precise enough and full 4:4:4.

Maybe someone could enlighten me?

4K is the future of all distribution... from YouTube to cinema. Trust me.

I do think you're right, but that's the future. I was talking status quo.

 

[Jannard]

Well, RAW does have it's advantages, like lower bandwidth and debayering choices, but i actually still don't understand the difference as far as color versatility is concerned. Both should give one the same control, assuming no funky in-camera gamma etc. adjustments being made, and the RGB being bit-precise enough and full 4:4:4.

Maybe someone could enlighten me?



I do think you're right, but that's the future. I was talking status quo.

Any RGB codec has baked in qualities that RAW does not...

Status quo... NO!

Jim

 

[Eki Halkka]

Any RGB codec has baked in qualities that RAW does not...

I guess i'm trying to find out what those qualities might be color versatility wise, as i really can't simply reason the answer - both seem to be linear RGB to me, only difference being the spatial arrangement of the color samples (which in this case would be superior in RGB).

I don't want to waste your valuable time on this, so i'll try to do some research and find some answers by myself (not that i would mind someone giving a detailed explanation here :smile5.

Status quo... NO!

Eh, like with "standard", i guess that too depends on the definition of the word - my intended one was simply "the current situation". The only point i tried to make was that most digital screens today are still 2K max. It will change of course, status quo is in constant motion... you being one of the forces that push it on.

 

[Graeme Nattress]

If you make RGB, it will have gone through some transforms that are not fully reversible back out to the RAW state.

When we know more about what is being done with this new camera, how it works, we can begin to talk about the pros and cons in more details. At the moment, all we can really say is:

Welcome to the 4k club,
What on earth do you mean by 16:8:8?????

Graeme

 

[Andrew M.]

I am bit confused. First they announced 17.7 MP sensor with 45deg pattern.

Next they announced 20MP sensor that looks like early RED 4K format with double the horizontal pixel count.

Than probably some marketing guy, totally hooked on HD format, tries to cook the simple explanation how good the sensor is by insinuating that the sensor would count as a 16:8:8 if it would be HD sensor.

This I understand but then don’t call it 4K. 2K sensor that is 16:8:8 is actually 4:2:2 once introduced as a 4K sensor.

Well good explanation for simple mind. But where this 20MP sensor popped up, after detailed explanation of 17.7MP one.

Are these two different projects, left hand doesn’t know what right hand is doing?

Or it is just project in progress so they can’t make its mind which way to go and they produce different prototypes?

This or other way looks like RGB is ruling there big time.

This is what happenes when you have 30 years of RGB technology behind you. Difficult to break the chains and get rid of it. Very painful process for large corporation.

 

[Graeme Nattress]

But chroma sub-sampling notation is NOT a sensor notation, and it use it with reference to a sensor is wrong.

Only thing we can do is turn up at NAB with zone plate t-shirts and stand in front of the camera and record the result.

Graeme

 

[Michael Olsen]

Only thing we can do is turn up at NAB with zone plate t-shirts

Available in the RED Store...when? :w00t:

 

[Andrew M.]

But chroma sub-sampling notation is NOT a sensor notation, and it use it with reference to a sensor is wrong.

Only thing we can do is turn up at NAB with zone plate t-shirts and stand in front of the camera and record the result.

Graeme

Do you think the marketing guys care about what notion it is.
To give him some credit, we don't know in what context he was using this 16:8:8 comparison.

 

[Graeme Nattress]

Well, maybe I'll send Jarred a zone plate image and we'll see if he can come up with one?

Graeme

 

[Eki Halkka]

If you make RGB, it will have gone through some transforms that are not fully reversible back out to the RAW state.

That's logical -

I'm still trying to wrap my head around this... are there some color affecting transforms in this conversion that can be adjusted by the user (like i.e. color balance), that can be only done at that stage, not after converting to RGB?

I've always thought (and also it's been my experience in practice) that if you have a high bit depth linear RGB file, adjusting that is very similar to adjusting RAW.

Sorry to keep this up, it's the little wannabe scientist / plugin writer in me that wants to know ;-)

What on earth do you mean by 16:8:8?????

That's indeed weird ;-)

 

[Eki Halkka]

This or other way looks like RGB is ruling there big time.

This is what happenes when you have 30 years of RGB technology behind you. Difficult to break the chains and get rid of it. Very painful process for large corporation.

Actually, it's not RGB, it's YCrCb, luminosity, red chroma, blue chroma.

Pretty much everything in the video world is based on that, be it DigiBeta, HDV or whatnot. Everything with a ratio number is something else than RGB.

 

[Graeme Nattress]

Each and every transform you apply will loose you precision in the data if you're going to an integer codec afterwards. That's why we get best results when we do our transforms in 32bit float and output to a 16bit file.

Some transforms can clip data - colour space transforms and white balance are notorious for that, and in that case, it's not precision loss, but a data loss. With RAW you can always go back and do the transform differently to avoid the clipping, and perhaps secondary correct the colours to keep the ones you want and get everything much better than if you'd let the camera do it.

I think it was RAW that really helped photographers adopt digital photography - JPEG was too much like a print for them, and they wanted the digital negative. Lots of other benefits to digital too - not least price and image quality, but RAW was a good part of the move.

Graeme

 

[Eki Halkka]

Each and every transform you apply will loose you precision in the data if you're going to an integer codec afterwards. That's why we get best results when we do our transforms in 32bit float and output to a 16bit file.

Don't the ASICS in high end cameras also work similarly, using high bit depths and floating point math?

Some transforms can clip data - colour space transforms and white balance are notorious for that, and in that case, it's not precision loss, but a data loss.

In my hypothetical scheme, color balance could be left at the sensor's native preset, i.e. 5600 K., as the edjustments would be done in post, and the color space transform would be to linear RGB - which is, as far as i know, essentially the same what sensors output...

With RAW you can always go back and do the transform differently to avoid the clipping, and perhaps secondary correct the colours to keep the ones you want and get everything much better than if you'd let the camera do it.

That of course only applies if you let the stuff clip... assuming a high end camera, headroom shouldn't be a problem, so one can expose to protect highlights (just like with raw).

I think it was RAW that really helped photographers adopt digital photography - JPEG was too much like a print for them, and they wanted the digital negative. Lots of other benefits to digital too - not least price and image quality, but RAW was a good part of the move.

That's of course true. RAW is also leaps ahead what all the current "video" cameras can do as far as the output goes, most record in subsampled YCbCr with REC 709 color space or similar, and it's pretty hard to get a decent linear RGB from that mess (with the exception of some high end cameras that shoot log, which is a pretty good starting point too).

But my (likely wrong) speculation was that the camera would shoot high bit depth (i.e. 16 bit) linear RGB, with a supersampling sensor... that would be a different case, much closer to raw in my understanding. Of course, the fact that they use a ratio when describing the output suggests YCbCr output, which is a lossy conversion and makes my point moot ;-)

If, and that is a big if, that number actually refers to oversampled 4:4:4, and (big) if the camera can save linear RGB files with high enough precision with no other processing than demosaicing and scaling, there shouldn't be any perceivable color flexibility difference, as far as i know.

Or am i wrong?

 

[Graeme Nattress]

I really don't know what other cameras do in their image processing chips. To get answers there on bit depths etc, you'd really have to ask them.

Even on non-clipped data a colour space transform can cause clipping on saturated colours.

The basic idea with RAW is to do anything that requires care and precision in the environment best suited to that - that environment not being on camera on set.

Graeme

 

[Eki Halkka]

Even on non-clipped data a colour space transform can cause clipping on saturated colours.

But in this case, the colour space transform would be from sensor output to linear RGB, if that can even be called a transform...

The basic idea with RAW is to do anything that requires care and precision in the environment best suited to that - that environment not being on camera on set.

That i do understand, of course. I also think i've sidetracked this discussion way too much, sorry for that. I just find this stuff fascinating, it's hard to help myself :w00t:

 

[Craig Parkes]

Its very interesting looking at this from the on site broadcast world point of view, where cameras themselves arent even white balanced, all the color correction is done by a ccu in the truck to get things looking the same. It may be worth looking at things from that point of view, that this is the sort of world sony knows, where each device is just part of a signal chain, where conforming a signal, in a live broadcast environment, is way more important than the ability to heavily alter that signal in post.

Doesn't excuse their strange marketing, but if you think of Sony's contribution to cinema cameras as one coming from a bunch of broadcast engineers, and the marketing coming from people who have to deal with that legacy without actually understanding how any of this stuff actually works, its easy to see where they may be a slave to two masters with conflicting goals.

Thats the great thing about Red, only one master, Jim Jannard. And fortunately for us he has a tendency not to believe other people whn they say something is impossible, and instead looks for geniuses who don't know what impossible means either.

 

[Andrew M.]

Maybe RED has one master but many people to please. On top of it the Master had this idea to make the camera modular.

Well, just in here you have a few hundreds if not few thousands of combinations and several different markets to satisfy.

That is why baking RGB at such early stage of image processing would be a big mistake.

RAW is probably the best suited for Cinematography where the quality and precission is the most important, broadcast people will be happy with good conversion in post from RAW to RGB that always could be improved in time to the perfection. And there are many other markets that RED camera is aiming for.

Most of them requires RAW