"The Truth About 2K...Future of Pixels" via Creative Cow

[Matt W.]

The Bayer design is, by all accounts, great, and 2k from Red should be every bit as good as, if not technically better than, 1080p from the Genesis (which is a striped sensor design that samples from 12 million photosites or something) or a 2k film scan. As the above image shows, there will be less aliasing and certainly there will be as much resolution.

But the Red in 2k crop mode, or a single sensor 1080p camera, won't hold up compared with any of the above, or even the EX1 (which has three 1080p sensors for a total of 6+ million photosites).

The company has been fairly open about this: the new 5k camera is designed to be sufficiently oversampled to make 4k post worth it. With good lenses, it may be worth it. The bias goes back to the difference between containers and acquisition; the Red acquires footage at 4k Bayer, but 4k Bayer doesn't quite "fill" 4k (neither does Super35 film; IMAX, however, does). Look at an image from a large format film camera or medium format back scaled down to 4k, then look at Red footage. Red's footage isn't as sharp. So that's what the whole hoopla, which is really pretty silly, is about.

This is all about marketing and counter-marketing. Notice Galt says "Red isn't true 4k," not "Red isn't even 1080p," and 1080p is the best anyone could claim the Genesis is. And, as I like to point out, based on the new Fuji Eterna Vivid 500T's mtf curves, it's arguably way worse than either and it's certainly much grainier, and yet everyone so far seems to love how it looks.

 

[Michael Olsen]

I just wanted to make a quick note here: I'm not trying to disparage anyone, any specific resolution, or even compare cameras to each other. What I'm interested in is how two fields (photography, cinematography) evolved in a similar way (film -> digital), but with different paths.

It seems like photographers who switched to digital more or less jumped right over into Bayer CMOS and have been happy ever since.

Cinematographers got all sorts of different options which did all kinds of different things to achieve an image, and now there is dissent as to the right way to get an image off a piece of silicon...though I think, at least for the time being, that dissent is fading into a quiet acceptance of the state of the art.

 

[David Rasberry]

I see a lot of over enlarged "fine art" digital photographs that show significant color moire in fine details due to aliasing. The still guys seem to not care so much. Perhaps since it isn't moving and flickering, it is not as visually annoying, but it still bugs me.
I'll take a little softer image and a good OLPF any day.

 

[Graeme Nattress]

1) are any deficiencies in Bayer pattern cameras real or imaginary? Maybe stills photographers don't complain as there's no real issue to complain about.

2) bias through exposure to rather cheap consumer low resolution bayer pattern cameras

3) not that I know of for certain, but I think Nikon may had something like that in their earlier cameras.

4) You've got to measure it. F35 does 4:4:4, but the chroma is all offset at high spatial details through the stripe pattern. You're recording 4:4:4 but does the image count as that??

Graeme

 

[Jeff Coatney]

I have enormous respect for Panavision and John Galt, but I've heard nothing from any of John Galt's presentations that I could characterize as unbiased. I believe he is very bright, he knows his technology, but like many of his peers, he also has a vested interest in what he presents, how he presents it and most tellingly, what he omits from those presentations. Take from them what you will - make up your own mind.

 

[Michael Olsen]

1) are any deficiencies in Bayer pattern cameras real or imaginary? Maybe stills photographers don't complain as there's no real issue to complain about.

2) bias through exposure to rather cheap consumer low resolution bayer pattern cameras

3) not that I know of for certain, but I think Nikon may had something like that in their earlier cameras.

4) You've got to measure it. F35 does 4:4:4, but the chroma is all offset at high spatial details through the stripe pattern. You're recording 4:4:4 but does the image count as that??

Graeme

Thanks for that, Graeme. Sounds like we are on the same page. :beer:

 

[Blair S. Paulsen]

I have enormous respect for Panavision and John Galt, but I've heard nothing from any of John Galt's presentations that I could characterize as unbiased. I believe he is very bright, he knows his technology, but like many of his peers, he also has a vested interest in what he presents, how he presents it and most tellingly, what he omits from those presentations. Take from them what you will - make up your own mind.

+1

Cheers - #19

 

[Jesse Korosi]

When looking at the option of being able to punch in as much as you can on a shot when having a 4K image for a 1080 output it seems the "higher" resolution does indeed make a difference. For a lot of the long format shows I have worked on it seems common to use a wide or medium shot and punch in for a tighter shot in post rather then going back out and spending more money on shooting inserts if production didn't get the coverage originally...

 

[Bill Anderson]

Matt, it is not largely marketing: if you can't tell the difference between a 10 MP and a 20+ MP then you have not enlarged the prints to a degree that will ( undoubtedly) show the difference between the two (all things being equal). And since we're talking about the utmost in enlargements ( a movie screen or Imax) then we had better be very certain of how we compare things. There is little point in looking at 11X14 prints when you are preparing to shoot a 20 foot high backlit transparency.

I've heard this argument for a very long time. 35mm "looks as good as 4X5" and "4x5 is as good as 5x7" and "5x7 is as good as 8x10": ergo, 35mm looks as good as 8X10. Not so.

Most practitioners care about the end result and not the way it was derived. It means nothing to me if debayer uses voodoo to arrive at the final data, only that data and its subsequent quality matter.

 

[Robin Balas]

1) No they are not. Most live by "If it looks good, its good... To hell with theory!"
2) Maybe something connected to the fact that 1080p/2K is rather low-res compared to most high-end photography, and specially when done on digital MF equipment. And it is less visible in the normal sized prints in magazines compared to normal projection sizes - but that is only a guess from my side.
3) I can comment on what have been done with proffessional MF digibacks which come in 5 flavors or more,
- single shot bayer patterns with optional OLPF filters much like dSLR's and the RED Mysteriums.
- quad shot bayer patterns with piezo electric "motors" moving the CCD chip 1 sensel/photosite between each exposure (4 off) giving full color in each pixel. The 4.th reading provides an extra green reading which is used in some systems to check consistency in light levels and then adjust as it takes several seconds to shoot a seqence. This is tethered systems not capable of shooting without a attached Mac/PC to control the exposure and raw file merging.
- 16 shot bayer patterns like the one above, but with the possibility to move the CCD chip 1/2 a sensel doing a physical interpolation of resoluiton as well as improved color sampling. This provides more "real-life" resolution at least in theory. Personally I see very little benefit over upressing a 4 shot file in Photoshop. And that 16 shot mode is now removed in the newest digibacks supporting multishots.
- tri-shot monochrome CCD sensors with a motorised color wheel. Doing one full exposure per color without the fragile piezo mechanisms. This is outdated tech not in use very many places anymore.
- scanning backs using a striped CCD RGB sensor array enabling both 3 color sampling per pixel and a bayer like color interpolation but with a faster scanning time. This is used both in rotating panorama cameras and in flat field cameras like the Seitz 617.
Ref. http://www.hasselblad.se/media/1664002/perfectshape.pdf for more info on the IMACON multishot technology now owned by Hasselblad. I have an old CD-ROM somewhere with some raw files IMACON gave out on request from a studio shoot of some gold flaked designer shoes done single-shot, quad-shot and 16-shot showing a significantly improved color rendering from 1-shot to quad shot, but not really much more in the 16 shot. The CCD was a 16Mpx bayer chip.
All very fascinating stuff but not very relevant for the moving image. Note also that most MF sensors are CCD type, not CMOS.

4) No idea

I hope the RED 617 sees the light one day as it would be an amazing photographic tool as all digital 617 cameras (all 1 of them) are scanning backs preventing use of flash equipment.

MHO.

Thanks Matt and Graeme.

I now have one or two more questions...

1) If all of these professional still photography cameras use single sensor CMOS systems, are photographers as outraged about their lack of color response as some cinematographers seem to be about the use of Bayer CMOS in video/digital cinema?

2) Why is high MP Bayer apparently the gold standard for photography, but is something that people...look down on for motion picture work?

3) Have there been attempts at producing something akin to a striped+binning sensor for photography? Or the use of prisms? Though I imagine size would become a limiting factor in the latter case.

--

4) My question about 4:4:4 sampling...

What happens, say, if I am presented with a camera that will output 4:2:0, 4:2:2, and 4:4:4 to some very nice codec/format. What can I do to figure out how much color data is actually being recorded - one would hope there would be a difference in image fidelity between each level, but is there an empirical measure akin to MTF, but for color? Or is using MTF for each channel being recommended as a testing procedure?