RED B16 vs F23/Panavision

[Greg M]

On the other hand, I think there are much bigger problems. HDCAM for example is only 1440x1080 (and low chroma resolution and high DCT compression)... if anything, the "pseudo" label definitely applies. It is pseudo HD. Yet many shows are being shot and delivered on it.

I think you are confusing HDCAM and HDCAM SR. We have been discussing the SR format which is not 1440 X 1080, but full 1920 X 1080 and records all the info unlike its baby brother.

We can debate "psedo" or not all day long and it wont change the fact that SR is THE dominate format in the United States. Most of the post facilities have adopted SR as have the majority of the telecine facilities. The far majority of the networks accept SR for HD submissions, and many ONLY accept SR.

 

[M Most]

On the other hand, I think there are much bigger problems. HDCAM for example is only 1440x1080 (and low chroma resolution and high DCT compression)... if anything, the "pseudo" label definitely applies. It is pseudo HD. Yet many shows are being shot and delivered on it.

I have assumed through this thread that we're talking about HDCam SR, not HDCam. That limitation does not apply to SR.

It seems like Scratch would go hand in hand with converting Red footage onto SR @ 24Psf (4:4:4)? So I don't think it necessarily rules out SR/24Psf workflows?

But it does rule out simple live recording, as even if you happen to have a Scratch workstation on location, it doesn't do live pass through. Of course, unless some sort of debayer is being applied to the live output (and at this point I'm assuming it's not, because nobody from Red has said otherwise), it's the same as recording from live playout of the proxies - not quite the quality level expected for master level recordings. So the whole point may be a bit moot.

 

[Dj Joofa]

It just seems like the RED should have a lot more to offer in it's image creatively with the type of data it records vs F23 or Genesis.

Don't know if Red is applying a curve to linear data before wavelet encoding. I remember somebody saying that some of the other cameras do so before compression. However, if Red is not applying such curves before compression then it will be helpful to do so since in addition to reduced noise in darker areas (resulting from later inverse mapping the curve), it shall help to achieve less distortion for the same bitrate, and hence, a sharper image may be achieved.

I.e., compression applied to curve-applied data should have less distortion at the same bitrate vs. the same compression applied to linear data; or lower bitrate at the same distortion.

 

[Patrick Tresch]

If they apply a curve in the high lights to get more lattitude, they would diminish resolution as there would be more banding (less bit per light increase) when an inversed curve is applied. If the "highlights" are used as your neutral tones (like when you expose at 200 asa) you would have your midtones with less informations.

My 2 cents.

Pat

 

[Dj Joofa]

If they apply a curve in the high lights to get more lattitude, they would diminish resolution as there would be more banding (less bit per light increase) when an inversed curve is applied. If the "highlights" are used as your neutral tones (like when you expose at 200 asa) you would have your midtones with less informations.

My 2 cents.

Pat

The effect of a curve similar in shape, but not necessarily in formula to log, or a power function, will have very little effect in highlights. Highlights will not be much effected by the curve. Its the darker areas that are more effected.

Information preservation (distortion measure):
=============================

I would tend to think that a curve of the type mentioned above will increase the correlation between neighboring pixel values. An increase in correlation always results in the lower bitrate per sample for the same distortion; or less distortion for the same bitrate.

(BTW, increase in correlation is a reason that a higher resolution image (say 4K) image is more compressible in bits per sample compared to a lower resolution one (say 2K) because the correlation between neighboring pixels increases.)

Noise immunity:
===========

Many forms of noise can be modeled as additive if the following two conditions are satisfied:

(1) bitrate is reasonably high enough
(2) some underlying statistics are Gaussian (normal)

Under additive noise in darker areas, inverse mapping the curve will diminish the noise more than just added to linear data.

 

[Phil Bates]

Back to P vs PsF for a second. If I understand correctly, PsF applies to the way the signal is played from a deck or a camera. It merely derives two fields from one progressive frame and plays them as if they were interlaced. So, when a digital capture is done and you are looking at the discrete frames on a computer monitor, they are still basically P. So bottom line is that PsF converted to a QT movie becomes P. Correct?

Thanks,

Phil
http://www.artbeats.com

 

[David Wyatt]

Absolutely did not look like film.
Graeme

Apart from the bits shot on film...surely??

Personally I thought it was a great movie - [unfortunately we don't make many of those over here any more :-( ], which just just goes to show it's not all about resolution, numbers, maths, equations, computer science...yawn...

 

[Graeme Nattress]

Yes, the film bit at the end looked like film But it was one of worst movies (from a plot / enjoyment) POV I've ever sat through... But we're getting rather off topic. If we want to talk about this, let's move it elsewhere.

Graeme

 

[David Wyatt]

Definitely needs a new thread..."Graeme reviews movies" - I'd definitely read it!! ;-)

I had you down as a horror fan...maybe not a DV horror fan tho' :-(

 

[GlennChan]

I have assumed through this thread that we're talking about HDCam SR, not HDCam.

Sorry, I wasn't clear. I was just saying that there is a lot worse pseudo-ness than HDCam SR not supporting P. I kind of went on a digression there...

 

[GlennChan]

Don't know if Red is applying a curve to linear data before wavelet encoding.

It shouldn't be doing this as it's recording 12-bit linear.

Somewhat relevant thread:
http://www.reduser.net/forum/showthread.php?t=619
Graeme's reply there is probably what you're looking for, as Red has presumably tested both 12-bit and 10-bit (e.g. with log curve) and found that 12-bit works better for them with the scheme they are going with (which might involve some secret sauce?).

 

[Nova Invicta]

Mixed Points

Mixed Points

Jim
Others here including you state technology moves on so Panavision "could" use a CMOS sensor in its next camera with a bayer filter and bayer filters themselves are being improved by the likes of Kodak. CMOS sensors are improving but so are CCDs and application specific sensors will rule the day in the future.
Graeme
Jo public couldnt give a fig whether a film was shot on film, on HD or drawn in a computer as long as its what they want to see and it entertains something far too many people here forget. They certainly dont go to the cinema to compare picture quality or ask did it use a striped sensor or a bayer.

To the entire Red family live and let live its certainly makes either rich or poor people look small if the only way to make a point is to constantly knock what the other guy is doing, sure sell your product but I for one will never be a slave to a single system look where that nearly got us with Microsoft.

 

[sbcooler]

As a big fan of U2, I enjoyed the U2 3D movie very much and loved how it came out...although from my novice eyes, I did see some shots where they light was harsh and some highlights looked blown out...I saw that periodically...at least I think I did.

Nevertheless it was well shot I think. Again those Sony F950's are 2/3" sensors so a direct coparison to Red's Scarelet would be interesting...although comparing a $110,000 cam to a $3,000 cam is not quite fair, but I wonder if you could even compare it...if you can compare it somewhat, even a little bit unfavorably for Scarlet, that is a huge victory for RED. HUGE.

I do wonder if their would be much of a noticable difference purely in image quality with the U2 3D shoot as is, vs. if done with RED. A rock show is a demanding environment to shoot in. Especially when the band keeps their light show unaltered!

 

[cinepost35]

As a big fan of U2, I enjoyed the U2 3D movie very much and loved how it came out...although from my novice eyes, I did see some shots where they light was harsh and some highlights looked blown out...I saw that periodically...at least I think I did.

Nevertheless it was well shot I think. Again those Sony F950's are 2/3" sensors so a direct coparison to Red's Scarelet would be interesting...although comparing a $110,000 cam to a $3,000 cam is not quite fair, but I wonder if you could even compare it...if you can compare it somewhat, even a little bit unfavorably for Scarlet, that is a huge victory for RED. HUGE.

I do wonder if their would be much of a noticable difference purely in image quality with the U2 3D shoot as is, vs. if done with RED. A rock show is a demanding environment to shoot in. Especially when the band keeps their light show unaltered!

It seems like if U23D could have been shot on RED (B16) it would have allowed the producers a lot more flexibilty in post with the RAW files with all the different harsh colors and lights in the concert over a destructive format like SR. And if RED has 4 times the image data over SR. That seems like the film could in theory have held up MUCH MUCH better than SR footage blown up and projected on an IMAX screen. Also RED cameras (If available as an option when the film was made) would have allowed the cinematographers who shot it more creative options over 2/3" Cameras.

Not knocking the film at all I think it looked great....