Jump to content


Photo

2K v. 4K And Beyond


  • Please log in to reply
17 replies to this topic

#1 Joe Giambrone

Joe Giambrone
  • Basic Members
  • PipPip
  • 98 posts
  • Director

Posted 04 December 2008 - 06:54 PM

I am trying to assess the impact finishing at 2K will have on image quality.

Are there pros vs. cons here who would like to make their case? Is 2K more acceptable for certain situations (1.85 v. 2.40)? Or certain genres (action, horror)?

As computer technology explodes, the reality and feasibility of 4K DI will become more affordable and practiced. We just don't know when. Since it's already happenig with 4k digital cinema acquisition, then we really need to figure out where the trade offs are, and how today's digital limitations will be perceived in a couple of years when the public becomes accustomed to more resolution.

A 4K image has four times the pixels of a 2K image. Some say it isn't a "true" 4 times the detail however. Is that technically, strictly true, or does it depend on a lot of factors?

The 1920x1080P "standard" is ten year old TV technology. Yet, people are filming big screen features on it today. What is your opinion on how well this standard holds up against 35mm and higher resolution digital?

Opinions across the internet vary. Is there a consensus on these matters here?

Thanks for your consideration of the questions.
  • 0

#2 Joe Giambrone

Joe Giambrone
  • Basic Members
  • PipPip
  • 98 posts
  • Director

Posted 04 December 2008 - 07:05 PM

I forgot to ask is anamorphic finishing a way to get more resolution out of the 2K data? And is there some overview on a webpage out there on how you can finish to an anamorphic 2.40, even if you shoot spherical?

How can one get the maximum image detail out of a 2K DI process? What parameters should be set?
  • 0

#3 Paul Bruening

Paul Bruening

    (deceased)

  • Sustaining Members
  • 2858 posts
  • Producer
  • Oxford, Mississippi

Posted 04 December 2008 - 09:48 PM

Valid opinion, yes. Consensus, no.
  • 0

#4 Elliot Rudmann

Elliot Rudmann
  • Basic Members
  • PipPipPip
  • 208 posts
  • Other
  • Chicago

Posted 11 December 2008 - 12:59 AM

Hi Joe,

Working at a DI house in Chicago, I think I can sufficiently answer a few of your questions. There are many considerations to take into account when deciding to scan at 2k vs. 4k. I would say that the majority of feature films and commercials shot on 35mm (at least in Chicago) are scanned at 2k or HD resolution (or scanned at 2k, then down-rezzed to 1920x1080 HD) simply because you get great quality with cost-efficiency. And this is regardless of the aspect ratio. You're correct in that with 4k, you do gain a noticeable amount of detail, how much is difficult to quantify for a number of variables like film grain, focus, etc.

Even with the most powerful computers, 4k footage is extremely a difficult, if not impossible, workflow to deal with on a personal level. This is something a lot of RED camera users just can't seem to grasp - yes the camera can shoot 4k, but it's nearly impossible to finish 4k without a post house intervening. 2k is still the most manageable (professional) resolution to deal with today, so it's what most companies deal with.

The DI house I work at uses an Arriscan, which can scan at 2k, 4k, and even 6k. We've never scanned an entire project at 4k. We have scanned 4k selects for shots that were going to have heavy VFX rendered over them, and other times we have scanned at 4k because the client wanted to significantly enlarge a portion of one shot without making the image too soft. For our Arriscan, it takes 1 second to scan 1 frame of film at 2k resolution. It takes 4 seconds to scan 1 frame of film at 4k or 6k resolution - clients factor this in when they decide whether or not they way to pay us 4x the amount for scanning time...guess which one they usually end up choosing?

The genre of the film usually makes no difference for how the film is scanned - again, it's the budget of the production that determines it. I guess if a certain genre like sci-fi where there might be lots of CGI and visual effects may end up scanning more footage at 4k than 2k - but again - that depends on the production.

It isn't so much a hunger for resolution as it is a desire to get the best possible final product for a reasonable cost. We have done 2k film-outs for features that look great on the big screen. For most producers, that's enough.

"How can one get the maximum image detail out of a 2K DI process? What parameters should be set? "

Well, a great scanner helps. I'd say the Arriscan does a phenomenal job, because it can also self-calibrate its settings to correspond with the type of film stock that is being scanned, allowing it to pull out the maximum tonal range and right color balance for each specific stock. It outputs 10-bit log dpx files which give colorists a great deal of range to work with. It doesn't offer real-time scanning like a Spirit 2k - but I've seen examples and talked to enough colorists to know that 2k images do look better on the Arriscan.

I think it will take a bit of time before we move beyond the hi-def "standard" - I put that in quotes because it still hasn't standardized for many popular channels (MSNBC is just starting to upgrade in a few months). Broadcast companies make considerable financial investments to go the 1080p route...they're going to stick with it as long as they can while others still catch up.

Hope this helps clear up a few things. Let me know if you have any more questions. My advice in a nutshell: finish on 2k.
  • 0

#5 Joe Giambrone

Joe Giambrone
  • Basic Members
  • PipPip
  • 98 posts
  • Director

Posted 18 December 2008 - 02:16 PM

Hi Elliot,

This is good stuff. Thanks.

Does the anamorphic finish use more of the negative on the final print? I was trying to figure out which aspect ratio would provide the best image, and I had assumed 1.85. But, the anamorphic, with an anamorphic lens on the projector looks to use more of the negative.

As for doing a DI on an anamorphic scan -- is it desqueezed to fit into 2048 x 8xx frame size? Or is the scrunch left in and the maximum frame size used (2048x1440?)?

If it's the former, I sense a big loss in quality. If the latter, it looks like the best possible quality you can get out of 2k.

How does it actually work?
  • 0

#6 Elliot Rudmann

Elliot Rudmann
  • Basic Members
  • PipPipPip
  • 208 posts
  • Other
  • Chicago

Posted 08 January 2009 - 10:17 PM

Hey Joe ?

Good question, relatively complicated answer? skip to the last two paragraphs if you don?t want to be really confused J

Before choosing between whether to shoot anamorphic/super 35/flat (1.85), you should ask yourself what your final output is going to be? Will you be printing back to film? Or will you end up needing an HD-master for television and/or DVD release? Or all three?

It?s very rare for us to receive films that were shot with anamorphic lenses go through the DI process and end up going back to a 2.35 film out. Shooting with 3-perf super 35 gives you virtually the same image quality when going through the DI process to a 2.35 finish and potentially saves thousands to tens-of-thousands of dollars in budget (anamorphic lenses cost considerably more to rent than spherical ones for S35, plus savings from shooting 25% less film stock with 3 perf). Basically, it?s rare to see someone shoot a project anamorphically with the intention of doing a DI finish.

Many old school DPs love the look of an optically printed/contact printed anamorphic print, and while I?ve never seen a comparison between a 2.35 film print made from a DI anamorphic print and an optical anamorphic print, my personal opinion is that the optical print would look better. A 2k scan just wouldn?t do it justice.

Nevertheless, if you were doing a DI to film-print for an anamorphic project, we would scan the entire frame, (2048x1556- stretched/unsqueezed), then color grade it in the proper squeezed format (2048x8xx), and print back to film with the unsqueezed frame, so that the projectionist would de-squeeze it with the proper lens, haha ? don?t think I?ve ever used ?squeezed? so many times in one sentence.

So if you?re doing a film-out from a DI anamorphic project (intended for 2.35 projection), it will look better than a 1.85 DI finish on film because there is simply more image space on the anamorphic negative. Anamorphic cinematography uses the full frame of the 35mm image, Super 35 and Super 1.85 do not. So you will get more resolution out of a scan from a negative shot anamorphically, but even then, a 2k scan of it will barely edge out the 2k scan of a Super 35 negative; that being the case, why even shoot anamorphic? Think of it this way; 2.39 anamorphic more of a *look* whereas Super 35 is simply an aspect ratio.

If all you?re going to do is end up with an HD master after the DI, then the difference in image quality between shooting anamorphic/Super 35/Super 1.85/ even academy 1.85 will be virtually indistinguishable. If you?re going to a release print, you need to figure out how you want it projected; 1.85 or 2.35? If 2.35, an anamorphic 2k DI will look all right (not as great as a contact print or optical finish), and a Super 1.85 2k DI will look better in 1.85 projection and 2.35 projection than an academy 1.85 negative would (more negative is used in Sup1.85). And a 3 perf Super 35 2k DI will look virtually the same as a 2k anamorphic DI in a 2.35 film projection with maybe a very slight decrease in image quality because S35 doesn?t use the whole frame. See this website for info on different aspect ratios: http://www.efilm.com/aspect_ratio.php

Assuming that you lit and exposed your film well, I believe that image quality in the DI process is determined more by the scanner/telecine machine and the colorist than the aspect ratio in which you shoot. I feel that 3 perf Super 35 format is the most practical way to go through a DI if you want a film release print projected at 1.85 or 2.35 because most 35mm projection formats (virtually everything but anamorphic) don?t even use the 4 perforation height of the 35mm frame. To echo what David Mullen says: It?s a matter of look. Check out what he says here: http://www.cinematog...showtopic=10982
  • 0

#7 Phil Rhodes

Phil Rhodes
  • Sustaining Members
  • 11939 posts
  • Other

Posted 09 January 2009 - 05:12 AM

I think it's worth mentioning the fact that merely converting between digital and photochemical formats is lossy. You could take a piece of film which has sort of 2K-ish worth of data on it and scan it to a 2K file, and you wouldn't still have 2K-ish worth of data. Conversely, I've always felt that if you shoot HD/2K and maintain it pixel for pixel through your post chain to a HD or 2K projector, you're probably looking at a result that's as sharp as most 35mm print exhibition. Take film and scan it, and project it digitally, and you lose. Take your digital original, scan it out to film, and project it, and you lose. This is true regardless of pixel counts. The lossiest situation of all is a DI because you're often converting twice, so there's not much doubt that using higher resolutions for the intermediate helps.

The point is that what's "good", "better", or "as good as" is very dependent on circumstance and the "good enough" numbers for a situation like DI do not necessarily inform what's "good enough" for a purely digital pathway.

Finally, there are many situations in which 35mm o-neg does not resolve even 2K. I have, memorably, worked with oscar-nominated material which was shot super-35 on 500 speed stock, and I would challenge any idea that there was really more than about a K and a bit of information in it.

P
  • 0

#8 Elliot Rudmann

Elliot Rudmann
  • Basic Members
  • PipPipPip
  • 208 posts
  • Other
  • Chicago

Posted 09 January 2009 - 06:18 PM

Absolutely Phil, I do believe that 2k can sometimes be overkill for the higher speed 35mm stocks, but the lower grained ones like 5212 and 5201 I feel deserve it (especially if exposed well and shot with sharp lenses). I do feel that 2k scans (even from 500 ISO stock) from our arriscanner hold its own against HD video and especially 2k RED material, which I find a bit too compressed for up close viewing, even on our 2k projector. We recently scanned in a commercial shot on 5212 with master primes and, when projected (from our digital projector) I was quite amazed, although this project won't get a film release so I can't say for certain it would look better than an HD video digitally projected. But like you said, it's the actual material that is filmed (or recorded) that makes the ultimate difference in how much can be truly resolved.
  • 0

#9 David Mullen ASC

David Mullen ASC
  • Sustaining Members
  • 19760 posts
  • Cinematographer
  • Los Angeles

Posted 09 January 2009 - 06:39 PM

Again, you want to separate measurable resolution from ideal scanning resolution, which should be a little higher than the measurable resolution.
  • 0

#10 georg lamshöft

georg lamshöft
  • Basic Members
  • PipPipPipPip
  • 310 posts
  • Berlin

Posted 10 January 2009 - 08:25 AM

Scanning a 24.9mm wide negative with 2048 pixels means scanning @ 2090ppi! I don't think that comes even close to the information density of modern film stocks!? High resolutions also result in higher contrast at lower frequencies!

From my experience, about 4000ppi are necessary for modern slide film, scanning at 6300ppi or even 8000ppi creates very large files with lots of grain (at 100% they just look horrible). But seeing the grain/film structure more precisely seems to be important for filters (noise/grain reduction). I filter them, then resize the files and sharpen the image to the given output size.

Are these comparisons from ARRI unrealistic (downloadable as uncropped TIFF at their ftp-server,ARRISCAN @ 2k left (sized up to fit 4k), 4k right): ?

Attached Images

  • 2k4k.jpg

Edited by georg lamshöft, 10 January 2009 - 08:28 AM.

  • 0

#11 K Borowski

K Borowski
  • Basic Members
  • PipPipPipPip
  • 3905 posts
  • Camera Operator
  • I.A.T.S.E. Local # 600 Eastern Region

Posted 10 January 2009 - 09:33 AM

I'll go with 4K! :D

Now, since you're talking about stills, which are usually what, twice the size or 2 2/3 times the size, or are you using a 3- or 4 perf. area equivalent of 2K and 4K?

Obviously a 4K of a larger piece of film isn't the same as a 4K scan of a smaller frame.

I think that 3.2K (or whatever the naming convention for the midpoint between 2- and 4K is today) should be the minimum, 4k would be better. I can even see a big difference in 4K film scans on my HDTV. Used to think it was the fault of the 500-grain film, but the problem of noise disappears on 4K scanned material.
  • 0

#12 David Mullen ASC

David Mullen ASC
  • Sustaining Members
  • 19760 posts
  • Cinematographer
  • Los Angeles

Posted 10 January 2009 - 12:05 PM

The 36mm-wide FF35 still frame is 1.5X wider than the 24mm-wide S35 cine frame.
  • 0

#13 georg lamshöft

georg lamshöft
  • Basic Members
  • PipPipPipPip
  • 310 posts
  • Berlin

Posted 11 January 2009 - 05:23 AM

That's why I convert resolutions like 2k/4k into ppi (pixels per inch):

2k for a 24.9mm wide S35-frame -> 2090ppi
4k -> 4180ppi

A 2k DI from a 35mm still-slide/negative (24x36mm) would result in a file with less than 6 megapixels! No professional would scan at these low resolutions - why should cinematography go this route when they're willing to pay for 35mm in the first place? To offer a properly scanned, "remastered" special edition in 20 years?

I think the differences in BluRay-IQ are unbelieveable. Many films made with the best equipment available look like crap: unsharp, noisy (not grain but scanner noise and grain aliasing) and were most likely transfered from the dailies from a telecine (?)... that makes it quite easy for HD-cameras...
  • 0

#14 Michel Hafner

Michel Hafner
  • Basic Members
  • PipPipPipPip
  • 300 posts
  • Other

Posted 18 January 2009 - 09:16 AM

I think the differences in BluRay-IQ are unbelieveable. Many films made with the best equipment available look like crap: unsharp, noisy (not grain but scanner noise and grain aliasing) and were most likely transfered from the dailies from a telecine (?)... that makes it quite easy for HD-cameras...

Can you give some examples?
  • 0

#15 georg lamshöft

georg lamshöft
  • Basic Members
  • PipPipPipPip
  • 310 posts
  • Berlin

Posted 18 January 2009 - 06:53 PM

"Finding Neverland", "Ocean's Thirteen", "Crash"... They're not uprezzed DVD-transfers but they don't do the original justice, simply not very 35mm-like. Scanned properly with ~4000ppi and then downsampled to 1080p (to avoid grain alaising) even grain without filtering shouldn't be very obvious with slower stocks on Super35.

Attached Images

  • noise.jpg

  • 0

#16 georg lamshöft

georg lamshöft
  • Basic Members
  • PipPipPipPip
  • 310 posts
  • Berlin

Posted 18 January 2009 - 07:03 PM

That's a 100% crop (JPG-conversion with max. quality in PS) from a 2008 Blu-Ray from a 35mm-film (bright daylight-interior):

Maybe it's the scanner-operator, but to me the scanner (quick&dirty 2k telecine?) is crap and should only be used for dailies, not BR-transfers/DIs...

Actually the ESGR-Super16-examples look better...

Edited by georg lamshöft, 18 January 2009 - 07:06 PM.

  • 0

#17 Elliot Rudmann

Elliot Rudmann
  • Basic Members
  • PipPipPip
  • 208 posts
  • Other
  • Chicago

Posted 19 January 2009 - 12:29 AM

Maybe it's the scanner-operator, but to me the scanner (quick&dirty 2k telecine?) is crap and should only be used for dailies, not BR-transfers/DIs...


I couldn't agree more. 2k telecine is no match against a 2k scan from a dedicated film scanner. This is highly evident in wider shots, where more noise is apparent in clean, blue skies and edges are often oversharpened. Scans from the Spirit 2k seem to fall victim to this.
  • 0

#18 Michel Hafner

Michel Hafner
  • Basic Members
  • PipPipPipPip
  • 300 posts
  • Other

Posted 21 January 2009 - 05:53 AM

"Finding Neverland", "Ocean's Thirteen", "Crash"...

I haven't seen these myself yet but Neverland from DI looked already grain filtered to me in the cinema, Crash is a 'low bit rate' MPEG2 compressed disk that got a lot of flack for compression issues and Ocean's Thirteen looks apparently weird anyway because Soderbergh went for a special look. That does not mean that all of these can not also have grain aliasing issues as well.
  • 0


Aerial Filmworks

Willys Widgets

Paralinx LLC

Broadcast Solutions Inc

CineLab

FJS International, LLC

Gamma Ray Digital Inc

Glidecam

Ritter Battery

Visual Products

Wooden Camera

Media Blackout - Custom Cables and AKS

Opal

CineTape

Rig Wheels Passport

The Slider

rebotnix Technologies

Tai Audio

Technodolly

Metropolis Post

Abel Cine

Visual Products

Wooden Camera

Opal

Paralinx LLC

CineLab

rebotnix Technologies

Broadcast Solutions Inc

Media Blackout - Custom Cables and AKS

Ritter Battery

CineTape

Metropolis Post

Glidecam

Abel Cine

Gamma Ray Digital Inc

The Slider

FJS International, LLC

Tai Audio

Aerial Filmworks

Technodolly

Rig Wheels Passport

Willys Widgets