The truth about 2K,4K and the future of pixels

[Tenolian Bell]

Well 4K in the world of the professionals who do this, and you say “4K,” it means you have 4096 red, 4096 green and 4096 blue photo sites. In other words. You'll have 12 million green photo-sites, 12 million blue photo-sites, 12 million red photo-sites. That's 36 million photo-sites. A 36 mega-pixel image is what you get from a 4K scan."

 

"Now you know very well that you cannot take a 8.3 million pixel sensor and create 36 million out of that without interpolation. You are up-converting, and there's really no value to the up-conversion. There's no new information."

 

"So 4K is not these 8 mega pixel or 9 mega pixel or 10 mega pixel CMOS images for the Bayer pattern where they add up all the pixels in a row and say hey, we got 4K. The great perpetrators of that mythology have been RED and Dalsa. That's why I call these “marketing pixels." It's intentional obfuscation. But somehow the world has accepted that that's 4K. It's purely semantic."

 

"Now if you use the same arithmetic that these people are claiming they're 4K cameras are using, then Genesis would be 6K. Because it has 5760 pixels on one line: 1920 red, 1920 green and 1920 blue. But isn't that a little bit nonsensical?"

 

"I think that people will start to understand this and realize that it creates a terrible problem with post, because you have so much more empty data to process."

 

John Galt, Panavision

 

Creative Cow Magazine

[Phil Rhodes]

Ain't I bin sayin' it?

 

I said this several years ago at the launch of the Dalsa Origin camera, to which Dalsa folks responded with "well, but, yeah, but, no, but...".

 

Origin is (was?) an unutterably fantastic camera but it wasn't really 4K. Red is an unutterably mediocre camera and definitely isn't 4K!

 

P

[Tenolian Bell]

Yes Phil, I knew you especially would appreciate that article.

[Paul Bruening]

Indeed, Phil. You have been commenting accurately on many topics in the time I have been goofing around on this forum. There seem to be a crew of guys, here, who speak from base sense and knowledge such as you, David, Hal, John and so many more that it's too many to list. The rest of us absolutely do appreciate it. Keep it coming. Keep it coming.

[Jase Ryan]

And what of the Panavision Genesis? How does this relate to this conversation?

[Tenolian Bell]

The Genesis outputs 1920x1080, it was never sold as a 4K camera.

 

His point is that the Genesis has full 1920 lines for Red, Green, and Blue.

 

While the REDCAM or Dalsa Origin claim to be 4K cameras but don't have full 4096 for every color.

 

And what of the Panavision Genesis? How does this relate to this conversation?

[Keith Walters]

Ain't I bin sayin' it?

 

I said this several years ago at the launch of the Dalsa Origin camera, to which Dalsa folks responded with "well, but, yeah, but, no, but...".

 

P

 

 

Well whaddya expect?

They have a whole division set up to develop and market this concept.

There are engineers, laboratories, management and admin people, cafeterias, car parks, etc etc.

So, at some point the Division General Manager is going to say:

"Look chaps, I really don't thing this concept is ever going to bring any worthwhile financial return. Why don't we give it a miss eh? I'll have Miss Crookshank write out cheques your termination pay and entitlements; if we hurry we can be out of here by 5PM. Make sure you turn out all the lights, and for heaven's sake, don't leave anything in the fridge!

"See you all down at the Ferret and Millstone for final drinkies, at say; 7 O'clock?"

 

Yeah, that's REALLY going to happen...

 

They're always going to say whatever needs to be said to keep the overdraft ballooning a bit longer until the either the company either goes bankrupt or the parent company pulls the plug.

 

Origin is (was?) an unutterably fantastic camera but it wasn't really 4K. Red is an unutterably mediocre camera and definitely isn't 4K!

(Dalsa rep stamps his Hush Puppies, hissing in vexation.)

"Mr Rhodes! You simply do not understand about algorithms.

"Once you have the right algorithms, input resolution is completely independent of output resolution! Algorithms, algoritnms, algorithms!!

"This has been unequivocally established by years of discussion on Internet forums, countless pages of manufacturers' press releases, interviews by non-technical journalists with third rate-producers who most of the general public have never heard of (or seen their films), vast amounts of testimony from talentless non-entities who absolutely know, know I tell you, that the only thing separating them from making another Lord of the Rings is the Great Celluloid Conspiracy specifically engineered by geriatric silver-fixated old farts terrified, terrified of digital-savvy upstarts like them."

(Sighs noisily)

"What more evidence do you need?!!"

 

To be fair, I don't think the RED is all that bad, it's just if they had actually set out to actively inhibit its mainstream industry acceptance, I doubt they could have done a better job it :)

 

If they had concentrated more on making the thing the amiable industry workhorse that it is capable of actually being, instead of the inbred tempremental thoroughbred that it can never be, their market acceptance would have moved well past the noisome mess of music videos, cheap commercials and vanity projects enthusiastically festering away in that vast decomposing pus-infested toxic waste dump known as the US TV industry.

 

(The above assessment is entirely based on the one (1) TV movie I have actually managed to see that was shot on the RED. And no, I am NOT going to go out of my way to experience the RED experience; if the programmes don't come to me, well, says it all really....)

[Phil Rhodes]

(Dalsa rep stamps his Hush Puppies, hissing in vexation.)

"Mr Rhodes! You simply do not understand about algorithms.

 

You are so close to being precisely correct, down to the choice of both words and footwear, that I nearly snorted hot chocolate all over the keyboard.

 

P

[Keith Walters]

[/i]You are so close to being precisely correct, down to the choice of both words and footwear, that I nearly snorted hot chocolate all over the keyboard.

 

P

I have found a waterproof keyboard an excellent investment for those who spend significant amounts of time on Forums like this one.

 

Actually it sort of reminds me of a job I had many years ago, where one of our clients had a small video production facility located in a research psychiatric institution.

 

All of the cameras were mounted on the ceiling, with rain covers permanently attached. All the monitors were mounted in cutouts in the walls with 1/2" perspex shields dynabolted to the brickwork. Operation was only possible via remote controls, of which my boss kept a large supply of spares as the turnover was unbelievable.

 

The control room and VTRs were similarly located behind a 1/2" perspex screen, and all verbal communication via an intercom mounted in the ceiling.

 

Yet, none of the tapes I ever say showed anything but harmless-looking although obviously mentally unwell patients. God knows what must have gone on when the tape wasn't running...

[Karel Bata]

Interesting article. What made me really sit up was some discussion of HDR photgraphy, followed by...

 

"Today, that's only available in the still photography world. DynaMax is designed to do that for moving images. With those 6 red, 6 green and 6 blue photosites for each output pixel, you'll have the equivalent of shooting 6 images with different exposures at once, and blend them together to create a single high dynamic range image. You'll be able to capture extreme highlights, the near highlights, the mid highlights...."

"The DYNAMAX-35 sensor is a multimode video sensor capable of operating up to 120 fps at 6x HDTV and 30fps at full res of 37Mpix"

"In DynaMax we can control these individual photo sites so they have a short exposure, longer exposure and so on. So we can then take those exposures and blend them together to create a high dynamic range image, just as if you were shooting half a dozen different exposures."

 

Wow. Anyone know of any HDR videos viewable anywhere from this thing? Or is that something we'll have to wait for?

[Andreas Wessberg FSF]

I´m really bad with all this technical stuff.

But I get it, Red is not 4K. But how many K does the camera record then?

3K 2K or just 1K?

Can someone of you please explaine how to do the maths.

 

Andreas

[Phil Rhodes]

Two and a half to maybe almost three, depending on your point of view. What they're doing is arguably a misuse of terminology (which Silicon Imaging and Dalsa also do, or did). Traditionally, the "K" suffix was used to indicate thousands of pixels resolution from a film scan, which would by design have sampled RGB for every pixel Using it on a bayer sensor device which does not sample RGB for every pixel is highly questionable - usually, such things are measured in megapixels, as in a DSLR.

 

The difficulty of this is that there are a large number of variables involved, some of which are considered proprietary, so more or less all you can do is shoot tests. That's then difficult because you're also characterising a lens. The two principal opportunities for creative lying in bayer imaging are low-pass filtering and algorithms, the latter of which has become such a well-known weasel word that I nearly put it in quote marks.

 

The former refers to the need to low-pass filter - basically, blur - the image that lands on the image sensor, such that no fine details of it are smaller than the gaps between the pixels. This is complicated on a Bayer sensor because there are big gaps between pixels for red and blue, and smaller gaps between greens. The filter itself is, at the most basic level, just a ground glass placed an infinitesimal distance from the surface of the sensor, with the distance controlling the degree of low-pass filtering. Failure to do this correctly will result in one of two things - either more resolution shortfall than is strictly necessary, or higher apparent resolution with risk of aliasing, that crawling jaggedness in the image. Since the latter tends to look subjectively better in stills, and typically gives better numbers when aimed at a test chart, there's no prizes for guessing which option management likes best.

 

The second issue is the way in which the data from the sensor is interpolated to form three complete colour channels. There are a wide variety of techniques for doing this, from the extremely simple and artifact-riddled, to the extremely advanced, although at the end of the day there's no way around the simple fact that it is interpolation and you are guessing. It can be very good interpolation, directed by cross-channel assumptions about the saturation of real-world images, but it is unavoidably made up data. Beyond a certain point you can try so hard to do this that you will begin to introduce artifacts into the image*, and although Red is so heavily compressed it's difficult to evaluate this, it seems likely that the camera does suffer to a degree from overenthusiastic output processing because they were desperate to make it 4K when it just isn't.

 

Then of course there's compression. Red compresses at least 12:1, even if you don't consider the bayer mask itself to be compression. If you take that into account, it's something like 36:1. Either way, that's an extremely large amount of compression for an origination format. There have been repeated attempts to wave this criticism aside by mentioning the fact that they're using wavelet transforms (see JPEG-2000) as if that's some sort of magic bullet. Wavelets are management's dream - the artifacts produced by heavy wavelet compression are usually less visually obvious than those produced by DCT compression (see: JPEG, miniDV, MPEG, etc). While that is usually a good thing, it does mean that, since Red has no uncompressed output, we can't compare before and after. Therefore, the only people who can qualitatively evaluate the performance** of Red's codec are Red.

 

Further to this, while wavelets do generally produce visually unobtrusive artifacts (usually blurring), when they do fail, they fail gigantically and spectacularly in a way that is impossible to fix. The artifact in these circumstances usually looks like a chunk of the image has been scaled down and then back up again using a very naive algorithm. DCT codecs may produce blocking, but the location of those block boundaries is known and the problem is effectively addressed in most implementations, such as h.264.

 

Red may be a very sharp 2K camera and I wish they'd gone for that; instead, what they have is a rather questionable claim to a 4K camera.

 

P

 

* heterodyning additional information out of the image along various non-orthogonal axes would be that point, I think.

 

** wavelet-based codecs are notorious for creating images which have an amazingly high signal to noise ratio but which look like absolute junk, so the meaning of "performance" needs to be carefully evaluated here.

[David Mullen ASC]

I´m really bad with all this technical stuff.

But I get it, Red is not 4K. But how many K does the camera record then?

3K 2K or just 1K?

Can someone of you please explaine how to do the maths.

 

Andreas

 

The camera has a 4K Bayer CFA sensor. It records 4K RAW. That 4K RAW is converted to 4K RGB.

 

The confusion lies between the number of pixels being recorded and created in the final conversion versus measurable line resolution by pointing the camera at a chart. Generally a Bayer filtered sensor can only create about 2/3's or 3/4's of the original pixel resolution in terms of measurable resolution, hence why any 4K RAW recording will probably only measure out to 3K at best.

 

Not to mention that there is an optical low pass filter to soften high frequency detail to prevent aliasing, so you might not even want a 4K sensor to resolve 4K -- you'd be better off building a 6K sensor to get measurable 4K.

 

But the other issue is why is 4K this magic number? What else is always "true" 4K in comparison? Not 35mm film necessarily, it can vary. It's a useful goal, 4K, but it is also arbitrary. We scan 35mm film at 4K for RGB, but that doesn't necessarily mean that the negative will show 4K line resolution on a chart either, and especially not the same resolution for red, green, and blue. Generally you want to scan at a little higher resolution than the image itself has.

 

4K isn't really much of a measurement of anything other than the number of pixels you have horizontally, in which case, it is certainly accurate to call the RED a 4K camera in that sense -- it has a 4K sensor, it records a 4K RAW file, you convert it to 4K RGB... it's just not going to measure out to 4K line resolution when shooting charts. But we don't shoot charts, we shoot movies.

 

The real question is sharpness, detail, and aliasing artifacts, not "K".

 

I think there is way too much time wasted over this issue. Either a camera creates adequate detail for your needs in a pleasing, non-aliasing manner, or it doesn't.

[Andreas Wessberg FSF]

David and Phil, Thanks!

Always great and full information from you. Are guys superheroes?

 

Can´t say I totally understand everything but I get the basic.

I think I will continue to trust my eye and consentrate on the lighting and framing, leave the technical to

the ones who know.

0K or 6K, If it looks good then I´m ok.

 

andreas

[David Mullen ASC]

After using the RED and then the Genesis right after that, all the while looking at scans of the Super-35 feature I shot in Canada, my feelings about the whole issue have shifted. Truth is that resolution can be all over the map, shot to shot, no matter what format you shoot -- rarely do you achieve some sort of ideal situation where you are getting maximum image resolution. Even scene contrast can affect your perception of sharpness. And a shot that is slightly off-focus can quickly go below 1K in resolution.

 

After using the Genesis and getting the impression that the image was slightly "sharper" than the RED image, or crisper, despite being a 1080P recording (1.9K, or effectively 2K), but additionally, having more problems with aliasing with the Genesis than with the RED, I just feel that it becomes a textural debate as to what type of sharpness you prefer.

 

Now I have to admit that I haven't shot line charts of both cameras to have a basis of comparison in terms of what level of detail they are recording, I'm only going on perception.

 

But my impression is that Sony & Panavision generally opt for maximum sharpness out of the system at the risk of some aliasing, whereas RED opted to reduce aliasing problems as much as possible at the risk of some overall sharpness.

 

I go back and forth as to which I prefer, because at times, I'd say that the smoother, less crispy RED image feels more "film like" and less electronic than the Genesis image, but other times, I feel like the RED image borders on being a bit muddy at times.

 

The RED camera seems, to me, to have about a stop less dynamic range than the Genesis, which may or may not play into some perceptions of contrast and thus sharpness at times.

 

And as for the holy grail of film, I've had issues with the sharpness of the Super-35 photography I am scanning for efx shots right now -- I would have expected more sharpness. Some of it just has to do with shooting in such poor light conditions up in Canada on 500T stock and thus either getting an underexposed negative or shooting wide-open on lenses that need a little stopping down.

 

On the other hand, I just screened a print of "Astronaut Farmer", shot in 35mm anamorphic, scanned at 4K then downrezzed to 2K, and the sharpness was great despite my use of diffusion filters and smoke.

 

Makes me want to chuck all the other formats, film and digital, and go back to shooting in 35mm anamorphic.

[Guest Stephen Murphy]

Makes me want to chuck all the other formats, film and digital, and go back to shooting in 35mm anamorphic.

 

Amen to that brother.

S

[Sascha - Sash - Seitz]

Dear John,

 

the digital world of cinematography is still very confusing. Seeing all the new systems and technology rising every month, I just sick with the old fashion tools which have worked under the most extreme conditions.

 

Looking forward to hear from you soon.

 

Good Luck to you and Panavision.

 

Kind Regards

 

SASH

 

www.sashcam.com

 

Well 4K in the world of the professionals who do this, and you say “4K,” it means you have 4096 red, 4096 green and 4096 blue photo sites. In other words. You'll have 12 million green photo-sites, 12 million blue photo-sites, 12 million red photo-sites. That's 36 million photo-sites. A 36 mega-pixel image is what you get from a 4K scan."

 

"Now you know very well that you cannot take a 8.3 million pixel sensor and create 36 million out of that without interpolation. You are up-converting, and there's really no value to the up-conversion. There's no new information."

 

"So 4K is not these 8 mega pixel or 9 mega pixel or 10 mega pixel CMOS images for the Bayer pattern where they add up all the pixels in a row and say hey, we got 4K. The great perpetrators of that mythology have been RED and Dalsa. That's why I call these “marketing pixels." It's intentional obfuscation. But somehow the world has accepted that that's 4K. It's purely semantic."

 

"Now if you use the same arithmetic that these people are claiming they're 4K cameras are using, then Genesis would be 6K. Because it has 5760 pixels on one line: 1920 red, 1920 green and 1920 blue. But isn't that a little bit nonsensical?"

 

"I think that people will start to understand this and realize that it creates a terrible problem with post, because you have so much more empty data to process."

 

John Galt, Panavision

 

Creative Cow Magazine

[Andreas Wessberg FSF]

I won´t reject any camera or format. I welcome all new products.

Different tools for different projects.

Sometimes you need to shoot 35mm for the latitud and "beauty" and somtimes

hd because you need a higher ratio or just like the look for that project.

I just try to see all digital cameras and formats as different filmstocks.

 

andreas

[John Sprung]

I go back and forth as to which I prefer, because at times, I'd say that the smoother, less crispy RED image feels more "film like" and less electronic than the Genesis image, but other times, I feel like the RED image borders on being a bit muddy at times.

 

This is why I think the OLPF should be interchangeable. The DP should be able to choose the compromise point he/she wants. Much like you'd choose a film raw stock.

 

 

 

-- J.S.

[David Mullen ASC]

This is why I think the OLPF should be interchangeable. The DP should be able to choose the compromise point he/she wants. Much like you'd choose a film raw stock.

 

 

 

-- J.S.

 

Maybe, but I'm not sure a DP is qualified to make that kind of judgement, it sort of goes beyond a film stock choice, more of an engineering choice. It would be nice if the system were more "tweakable" in that regards, but it sounds like you're opening a can of worms...

[Bobby Shore]

4K isn't really much of a measurement of anything other than the number of pixels you have horizontally, in which case, it is certainly accurate to call the RED a 4K camera in that sense -- it has a 4K sensor, it records a 4K RAW file, you convert it to 4K RGB... it's just not going to measure out to 4K line resolution when shooting charts. But we don't shoot charts, we shoot movies.

I think there is way too much time wasted over this issue. Either a camera creates adequate detail for your needs in a pleasing, non-aliasing manner, or it doesn't.

 

thanks for that... I think it's easy to forgot that even though there's a highly technical aspect to being a cinematographer, at the end of the day it's just as subjective as any other art form. well said, David.

 

Bobby Shore

DP

LA/Montreal

www.bobbyshore.com

[John Sprung]

Maybe, but I'm not sure a DP is qualified to make that kind of judgement, it sort of goes beyond a film stock choice, more of an engineering choice. It would be nice if the system were more "tweakable" in that regards, but it sounds like you're opening a can of worms...

 

In a way, you're already making OLPF decisions when you choose diffusion filters. That's a better analogy as far as the look goes than raw stock. But changing the OLPF as an operational thing is probably more like choosing stock, you wouldn't want to do it as often is my guess.

 

Engineering can hang some numbers on aliasing, but what those numbers tell us about the look of the picture isn't particularly useful. How much aliasing to accept -- or even to choose as part of the look -- is a matter of taste. That makes it the DP's gig. Choice of shutter angle, for instance, is a case in which you're choosing an amount of aliasing due to undersampling. It's a case of shoot tests with everything the same except the OLPF, and compare the pictures. That's something you're qualified to do, very well qualified.

 

 

-- J.S.

[Phil Rhodes]

I think it's an engineering thing too, but it does raise the interesting possibility of placing the OLPF on piezoelectric buffers so as to be able to electronically control it, and probably to much finer tolerances than you could mechanically. I wonder if you could base the position of it on an analysis of high-frequency image components.

 

Of course, you'd have your image sharpness going up and down all the time...

[John Sprung]

Of course, you'd have your image sharpness going up and down all the time...

 

Yup, which I think makes it the DP's decision. It's taste, and the look of the show.

 

 

 

 

-- J.S.

[JohnSellers]

This certainly is an interesting discussion.

 

As I follow this thread there are two small things I hope you could clear up for me to help my understanding on this topic. Firstly, what is OLPF? as mentioned:

 

"In a way, you're already making OLPF decisions when you choose diffusion filters."

 

Also can someone provide me more information about aliasing? What is meant by this term?

 

Thanks in advance