Is R3D 4:4:4 or 4:2:2?

[Wayne Morellini]

Evin:

It has been my experience that ALEXA and F35 footage has "truer" skin tone reproduction than Epic footage. To me true color reproduction should allow for what is seen by eye on the set to be translated to what is seen on the screen. In order for this to happen a whole lot of technical crap has to happen first. There are steps to the process. Light energy needs to be captured. That energy needs to sampled and then converted (analog to digital) to a numerical value representing the intensity of that energy. This is then scaled into the degree of accuracy of the system: 8-bit 256 values, 10-bit 1024 values, 12-bit 4096 values, and so forth. At this point it is RAW and it can be recorded. This is what the Epic does and what the R3D file represents. But is is not an image. It is not what was seen on set. These numbers need to be crunched, algorithized, transformed, debayered, reborn, interpreted. Debayering is a computational act. It is math. And it does use "real" (discreet), sampled information but it also uses interpretive and calculated (non-discreet) information to turn the math into an image. My experience has been that the accuracy of this process is relative to how the image was discreetly sampled at the sensor. Math can be re-calculated to get slightly better results but math cannot add discreet values that were never present in the first place.

The Epic is an excellent 4K black and white camera. But it's color reproduction I find lacking. The idea that the end user must take on more responsibility and a greater understanding of RED's "mysterious" image-making process is simply silly. Yes, RAW image data can be computationally changed an infinite number of times without (theoretically) destroying the underlining data. But RAW image data cannot "magically" produce chroma resolution that wasn't there (sampled) in the first place. RGB, YUV, XYZ, or ABC do not really matter as they are simply quantitative spaces in which the RAW data will eventually inhabit. Just because I have a 16-bit RAW file or a 10-bit RGB image does not mean that I have all of the luma and chroma information that was originally present in the scene at the time of capture.

I have no interest in which camera system provides bigger and better technical results. My interest is in the system that provides results that are as close to what I see by eye.

BTW: if we say the ALEXA is 4:4:4 at 2K/HD and the EPIC is 20:16:16 at 4K then wouldn't your logic lead us to declare the EPIC is 4:2:2 at 2K/HD?

Which is exactly my point. The Epic does not see color like I see color by eye. The Alexa (for example) seems to mostly come dangerously close. We know the Epic has WAY more resolution in measurable luma and chroma. So what gives? If color stays the same at 4K and 2K/HD why build a 3K sensor (Alexa) to deliver HD? Or why build a 8K sensor (F65) to deliver 4K?

What Adrian says is true, for a lot of video work, correction is not a good option, you want something authentic looking now to pass on.

I was going to cut this down, but there are a few issues. Please refer to mupy previouse post yesterday or so. Technical accuracy is the only real good way to measure increased accuracy. Most cameras, from what I know, see poorly and then enhance the image to look better before recording or monitoring it. So Red, instead of making arbitrary choices, puts this raw footage out for peooke to adjust to hiepw they want, which gives them good flexibility, not only that they give range of performance apparently.

You can recreate missing descete values with Bayer. It is simply that luminance falls and color spreads in predictable ways across pixels, so are likely to be there is the missing ones. Real detail may change the color channel on a pixel and indicate changes to the other primaries at that pixel. Therefore (I remember I had some of these concerns 8 years ago) Bayer does a descent enough job recreating the missing colors for us. If you examine details accross frames, interframe debayering mentioned before) it is likely you can pick up the actual missing color as details move in the picture.

You can't just judge skin color by eye. I agree with many things you say about the problems they are fixing with skin tones, something funny there. It is like they are deliberately ignoring talking about it. But the fact is people's visual accuracy varies from person to person, from eye to eye, and over time, over the hour, mine does too. If you stuck me in front of a monitor to view Red footage, it is going to look different then monitoring on set, or in theater, simply because of relitive lighting, intensity and color shift in those environments. If you view Red footage on a non calibrated monitor under fluorescent lighting and view Alexia monitirung output on set lighting, they are likely going to appear to vary more than they really do. Another thing is that Alexa is giving you their look in monitoring, and you likely can recreate that reasonably well with raw epic footage.

The other aspect is that not only isn't our visual performance consistent, the camera has it's own standard in sensor capture and recording. Cameras generally actually see much more poorly in various ways than we do, unprocessed they look unacceptable. They use this recording standard to conform their image too (in Reds case it is a full range recording and image is set for finale recording standard for display) so it can be passed to displays accurately, which might rvmight not, confirm to the standard. Now the display often uses three binary color frequencies or three ranges of frequencies to reproduce them (hopefully accurately). But your visual system might not see this too accurately, so even though something seems accurate or not it might be the opposit. But, if we did not use the accurate technical standards, designed to deliver a good quality for the general population, the color would go further and further out along the cgain until it got displayed and it would statistically look a lot worse to a lit more people. As the guys indicated, checking the eyes is the best option if it appears wrong, and checking with everybody individually can give you a crude average under the conditions of the test (same monitor, picture, room lighting, distance) to see if they agree it is spot on to the original.

I would not call Epic a good b&w camera, it inky sees in color, and any color inaccuracy you are worried about will likely lead to grey scale innacuracies.

I feel like the question is getting semantically ignored while we all know exactly what he's asking.

Does the green channel have more resolution after debayer? Yes. In my experience blue and red are softer after debayer. ..

Should 4:4:4.. and tape all be relegated to the dustbin of history? Yes, please for the love of God yes.

Would that not depend on the individual type of debayering algorithm?

4:4:4 discete or ranges, no, that is part of authentic sampling. Tapes, my take design can take thousands of times more capacity than even hard disk (Common, there must be a business investor with some nouse around here somewhere).

Gavin:

You're getting close but I'm actually not interested in technical resolution whether it be luma or chroma. I'm interested in color accuracy, rendition, and most importantly beautifully natural skin tones. I constantly have issues with how the EPIC records and processes color. Especially skin tones. I find it very disingenuous that some among us would prefer to beat the 2-is-better-than-1 drum no matter what. More resolution (technical sharpness) does not always deliver superior results over fidelity (perceived sharpness).

Before muddying the water it is best to understand what real sharpness is and perceived. Because real sharpness will enhance perceived sharpness, so you still need it.

Double, per axis, resolution will enhance the number of color samples per half res pixel, increasing potential color accuracy.

Look Adrain, you've jumped in here and your comments are all over the place, and then you come out with "Actually the ARRI Alexa utilizes a 3K+ Bayer sensor which downscales to 2K/HD for proper 4:4:4 color resolution. " which you know is absolutely factually wrong. From there it's gone downhill rapidly.

The original poster asked what no-doubt they thought was a simple enough question, but in reality the answer is a hell of lot more nuanced than they anticipated. The more nuanced answers have been accurately given as part of this thread, but you're constantly wanting de-rail the discussion along your own lines and that doesn't help us out here. We've had some very good contributions to the discussion, but when you say: "but I'm actually not interested in technical resolution whether it be luma or chroma" you've basically just said you're wanting to discuss something entirely different to what this thread is about.

Graeme

I don't think he is too inconsistent and very the place, mistaken maybe. The problem is all this silly dancing around him, than explaining what you think he actually means and what things really mean, which should result in less and less dancing space. Addressing his questions on his terms, then he might see our terms.

Honestly, no. Among other things, I've yet to see an image processed through Redcolor3 and Redgamma 3 that includes a human face in closeup....

Yep, noticed that and asked for that, as ussual ignored this.

 

[Graeme Nattress]

OK. So when I find less than appealing skin tones from the Epic is that the result of the camera's colorimetry and not the RGB color fidelity issues inherent in a Bayer sensor?

Are you suggesting that over sampling chroma for a given target resolution (8K sample/4K target) creates has no discernible difference in a camera's color fidelity?

It does sound that you don't find the colorimetry appealing. I can understand you now. Please rethink your opinion when REDGamma3/REDColor3 has been released. It is based on a lot of DP and colourist feedback and represents a big improvement over what I've done before.

Oversampling in general, be it luma or chroma, is not a factor in camera colorimetry. I think that goes for most competently engineered cameras. The only situation where you'll get grossly visible colorimetry errors is in the situation of a line-skipping DLSR that renders false colours on high resolution detail - in other words chroma moire.

Graeme

 

[Evan Starkey]

I hope the colourist knew they where feed backing there own death warrants.......just kidding of course.......cant wait to see this.

 

[conrad gaunt]

I feel like the r1(m) I have handles skin tones pretty darn well.. and is extremely flexible in post anyway?

I kinda feel like the Alexa shooters are shooting to different parameters, I don't get the impression many are shooting ARRI raw , so they may have different expectations about colour calibration if they're recording Log. RGB streams that have been pre white balanced etc, but I'm not familiar with Alexa myself

With raw, for me anyway, its about capturing DR and positioning clipping levels and noise floor more than anything, I skew the preview image when the histogram says "let more in, or hang back"..and I don't get scared.. i do take care to position skin tones a little in the mix, and try to expose shots with a consistant methadology, but it all pretty elastic anyway .. that said, I never sharpen red footage, and I quite like milky HDR (not as in x) shots that generally pop up in RCX with the in-camera settings , and I sometimes struggle to have the desire to do the "crush all the blacks" thing.. I like soft

A downsample done in post is great for reducing any artifacts, whether compression, or debayer related, but it wants to be done in post. a 2x2 downsample to 1x1 will make things look a little softer, I sometimes think people are contrast addicts, but no point doing it in camera, post manipulations want to be 32bit IEEE, at least, including downsamples, so that you can reduce any (however minimal) compression artifacts, and not just tighten up the debayer

 

[Graeme Nattress]

I hope the colourist knew they where feed backing there own death warrants.......just kidding of course.......cant wait to see this.

Te He, that's funny.... But of course as a good camera only improves on the skill of the cinematographer, helping them achieve their vision, great colour science in camera just gives the colourist a better palette to work from.

Graeme

 

[Graeme Nattress]

Count the pixels and don't read marketing papers only. The F65 has 20 mpx vs. 14 for Mysterium X. That's less than 1/3 more. It doesn't help resolution if I tilt my camera 45 degrees, whatever Sony is saying.

And that 20mp is a marketing number too - when we say the MX has 14mp, that's pixels that can be used to make an image. The figure for the F65 is 17.7mp, not 20mp which is representing nearly 3mp of pixels that will never make it into an image. It's an over-inflated number. Same with the 8k. That is not a real figure. They only have enough pixels for 5.6k or via their rotated math 4k and no-more. For all of Sony's trumpeting about resolution, I've yet to see evidence of either low aliasing or resolution beyond 4k from their camera. If you get the chance, Uli, shoot me a zone plate on an F65!

Graeme

 

[Eric Lange]

Nice one Evan, I had a good laugh at that too!

Maybe in the future you have different famous “colorists” that you can pick from “out of the box”.

About zone plate… Ditto, Seriously Uli that would be amazing if you could; would probably be one of THE most important cinemagraphic “tests”/experiments of 2012.!!!! … Seriously…


How is it that “uncle Jim” has not rented or provided Graeme et al with an F65 to run through the mill?

The only advantage that I see with the photo site diagonal arrangement is the possibility of faster read out time perhaps for more broadcast oriented applications.

Are they really still peddling “new math” to sell the cameras?


I suspect they were originally worried that they went with photosites oriented at 45 degrees that the customers/punters would not like it or be able to get their heads around it as everybody knows “pixels” HAVE to be square and upright (right?). I think they devised this sort of marketing scheme to get people to go “AHHAAA” we must use 45 degree oriented photosites because it gives you MORE… MORE! (we want more, we always want more!).

More of what again ??? (wasn't it some fumbly mumbly numbers about this or that and a really nice green... something like that oh yeah the number 67 seems to be important wasn't it?).

Saying their sensor is literally “slanted” towards a more broadcast centric (fast chip read out time)… does not sound very “Cinematic”. I’m sure “they” feel rather pleased with themselves that they managed to make folks comfortable with the idea of photosites on a 45 degree slant. However it seems to be a bit of an “overcompensation” that usually stems from a deeper insecurity about something…

Testing and output is everything or at least it’s the only sensible start or basis for comparison (stating the bleeding obvious) would genuinely be fantastic to see real data and imagery from the F65. I want to know too!

[What is the daily typical rental rate for said F65 (if you can get your hands on one)?]

Cheers,

Eric

 

[David Rasberry]

I can tell you from decades of calibrating CRT projectors that regardless of how high the chroma resolution of your image is sampled, blue will still look softer than red and red softer than green due to the sampling limitations of human vision.
There are many valid reasons for the virtually universal acceptance of Bayer CFA sensors in digital imaging. It remains the closest and most efficient mechanical mimic of human visual perception to date. The filter dyes are not pure RGB, but modified to simulate the way the eye actually separates color information and the relative color sample ratios also mimic human color resolution discrimination. Nobody much complained about this technology until it was applied to motion cameras. Many other sensor formats and designs have been tried, but non has come close to being as successful for visually relevant results.
How the actual sensor data is processed is as much art as science, and still in its infancy compared to more than a century of film science.

 

[Uli Plank]

@ David: So right! If you ever had to calibrate a three-tube projector, you'd know about the difficulties of getting blue in focus on your retina…

IMHO it's quite telling that Sony gave up on three-stripe in favor of modified (i.e. slanted) Bayer pattern from F35 to F65.

Of course I'll test, but that may take some time. Hope the (relative) aliasing is not as bad as the F3's.

 

[Humberto Rivera]

Let me preface this by saying, I do not know much about what is being discussed here, just enough to be dangerous, it definitely sound interesting. Having said that we’re having a discussion on the eve of two important releases the Redgamma3, REDcolor3, and the “Red Dragon”. What we’re talking about are changes to the picture, past, present and future! From my way of thinking this is really important, as a Director I just want to make sure we’ve have the best possible picture now, and in the future, so when we’re dealing with Redgamma3, REDcolor3, Four, or Five, I can go back and look at DATA I recoded in my R3D FILES, and through Graeme’s magic, (it’s not really magic is the work of genuine genius) and retrieve the best images. Semantics to the four winds, thanks Graeme!

Humberto Rivera

 

[Graeme Nattress]

@ David: So right! If you ever had to calibrate a three-tube projector, you'd know about the difficulties of getting blue in focus on your retina…

IMHO it's quite telling that Sony gave up on three-stripe in favor of modified (i.e. slanted) Bayer pattern from F35 to F65.

Of course I'll test, but that may take some time. Hope the (relative) aliasing is not as bad as the F3's.

Indeed, after all the anti-bayer FUD, they were happy enough to use it in F3 and in their DSLRs, and now rotated through 45 degrees in the F65. Funny that....

Graeme

 

[Eric Lange]

I can tell you from decades of calibrating CRT projectors that regardless of how high the chroma resolution of your image is sampled, blue will still look softer than red and red softer than green due to the sampling limitations of human vision.


There are many valid reasons for the virtually universal acceptance of Bayer CFA sensors in digital imaging. It remains the closest and most efficient mechanical mimic of human visual perception to date. The filter dyes are not pure RGB, but modified to simulate the way the eye actually separates color information and the relative color sample ratios also mimic human color resolution discrimination. Nobody much complained about this technology until it was applied to motion cameras. Many other sensor formats and designs have been tried, but non has come close to being as successful for visually relevant results.
How the actual sensor data is processed is as much art as science, and still in its infancy compared to more than a century of film science.

Absolutely… it is well known in the field of so called Psycho-physics that small blue objects (especially at a distance) are hard to discern. This is entirely down to the composition of the Fovea/macula in the human retina of course.

Basically the S-cones in the retina responsible for detecting short wavelengths (i.e. blue) are much more widely spaced in the retina than the L and M cones. One of the reasons cited for this is that axial chromatic aberration of the human lens blurs the short wavelength portion of the retinal image. Therefore don’t evolve a density of blue sensitive cones that can’t be resolved anyway…. (right?).


There are a couple of ‘juicy” little facts about that.

1. In the very centre of the visual field in the central Fovea there appear to be NO short wavelength cones at all!

2. Secondly that this “spatial” defocus (of the blue/S cones) causes the short wavelength portions of an image to “code” neurologically slower temporal variations than L- and M cones especially for dynamic imagery. I.e. it’s a slower more blurry response and perception in the “blues”.

I think that might be an interesting thing to take advantage of when designing compression algorithms for dynamic sequences.

With the Bayer pattern, At least with the blue (of course the green is well taken care of) having a spatially offset blue “receptor” that has a more spatially generalized influence is completely consistent with how the human visual system operates. Actually it’s better than the fovea of the human eye… (I think).


E

 

[Eric Lange]

@ David: So right! If you ever had to calibrate a three-tube projector, you'd know about the difficulties of getting blue in focus on your retina…

IMHO it's quite telling that Sony gave up on three-stripe in favor of modified (i.e. slanted) Bayer pattern from F35 to F65.

Of course I'll test, but that may take some time. Hope the (relative) aliasing is not as bad as the F3's.

Uli that would be super awesome… In your own time mate!

If I worked for “team” F65 I would be shooting ONLY rectilinear resolution targets not circular ones.

It’s going to be interesting to see how far they go in calling an apple an orange by only very selective and particular “tests”.

E

 

[Hrvoje Simic]

Yeah, maybe those engineers @ Sony don't have a clue what they are doing.

 

[Graeme Nattress]

Hrvoje, Which is deliberately drawn with big gaps to make it look worse? Why are square pixels drawn as circles - perhaps to disguise that in reality, they're exactly the same - one just rotated through 45 degrees and uses pixels with half the area.

Graeme

 

[Les Dittert]

Reminds me of Horses teeth ....

Reminds me of Horses teeth ....

​

There's a story I'm rather fond of about Greek Philosophers. It goes thus:​

Three Greek Philosophers are dining at a roadside inn and arguing about how many teeth a horse has. The first one opines that it's 32, the second 34 and the third 36. They argue and argue as the wine is consumed, each coming forth with mathematical proof of why his and only his answer is the correct one.


Eventually, tiring of the argument, they beckon to an Arab sitting at the next table and ask that he hear each argument in turn and judge who is correct. The Arab listens patiently to each argument then says 'wait a moment gentlemen' and disappears. He returns a few minutes later and points to the third greek. 'You, sir, are correct'.


Although satisfied that the argument is finally over they want to know what abstruse reasoning the Arab used to determine the correct argument.


'Simple, sirs' he replies. 'I went out to the stables and counted them'.

So let's just worry about the output image!

 

[Martin Stevens]

The Output Image from the EPiC-X and Scarlet-X is SUPERB!

And, YES, the skintones are also SUPERB!

 

[Craig Weeding]

I agree with Les, let the pictures do the talking.

Both the RED One and the EPIC provide amazing pictures

The Sony may be a good camera but we travel a lot! and we are not interested in traveling around the world with cameras so much larger and heavier then an EPIC.

The Sony is a Beast, even if it has good pictures the price you pay in size and weight is too much for us.

We will take RED EPICs any day.

 

[Hrvoje Simic]

Hrvoje, Which is deliberately drawn with big gaps to make it look worse? Why are square pixels drawn as circles - perhaps to disguise that in reality, they're exactly the same - one just rotated through 45 degrees and uses pixels with half the area.

Graeme

Graeme,

Firstly, you have my condolences with Nth debate with Bayer and chroma over the years.

Secondly, its blatantly obvious it was drawn that way to accentuate the benefits. That's the marketing thing.

Disregarding that, and besides the difference in organic feel of each pixel layout,
analysis of the same surface area required to calculate one pixel shows spatially more symmetrical correlation of RGB chroma values.

 

[Gavin Greenwalt]

That was my point, that even with my 1/3rd pixel shift the differences become less noticable at higher resolution. So much so, that I contemplated how Bayer raw would look without debayering, Graeme?

I think 4:4:4 would still be somewhat important to special effects people Eric

More important to FX people is resolution and a lack of artifacts. I would rather have a moire free 2k than an artifact ridden 4k. And I would also prefer 16k:8k:8k to 4k:4k:4k. 4:4:4 is still important to me for 1080p but RED delivers that... so I'm happy.