Film-Tech Cinema Systems
Film-Tech Forum


Post New Topic  Post A Reply
my profile | my password | register | search | faq & rules | forum home
  next oldest topic   next newest topic
» Film-Tech Forum   » Operations   » Digital Cinema Forum   » 4K Digital Comparison (Page 1)

 
This topic comprises 3 pages: 1  2  3 
 
Author Topic: 4K Digital Comparison
Lyle Romer
Phenomenal Film Handler

Posts: 1331
From: Davie, FL, USA
Registered: May 2002


 - posted 01-20-2005 10:20 AM      Profile for Lyle Romer   Email Lyle Romer   Send New Private Message       Edit/Delete Post 
I had posted this in another thread but it seemed to get lost so I figured I'd make it a new topic:

I was just running some numbers to compare the resolution of Sony's 4K projector vs. 4K digital intermediates. I got most of this info from cinesite (except the super35 numbers, I got from bipack). So basically a 4K scan of a film shot in scope would be 3656x3112 (11.4 Megapixel), a film shot in super35 would be 3950x1647 (6.5 Megapixel), and a film shot in flat would be (interpolating from academy offset) 3656x1976 (7.2 Megapixel).

Sony's 4K projector is 4096x2160 (8.85 Megapixel). So would the following be correct to say (assuming they get the contrast and color to equal film at some point)?:

If a release print and a digital file were made directly from a 4k digital intermediate, the 4K digital projector will look at least as good if the source material was flat or super35 and only slightly worse if the source material is scope.

Since no release print is directly from a laser recorder, at least some of that resolution will be lost in practice. Is it then safe to say that with proper contrast and color, 4K digital projection should eclipse the quality of 35mm film projection.

I know that John Pytlak has commented before about the need for anti-aliasing filters which reduce the resolution of the fixed array projection system. Are these same filters requred when working with digital intermediates (or digital effects for that matter) when printed to film? Or does the film grain sufficiently randomize the image even though it is printed like a fixed array.

 |  IP: Logged

Paul Konen
Jedi Master Film Handler

Posts: 981
From: Frisco, TX. (North of Dallas)
Registered: Jun 99


 - posted 01-20-2005 11:56 AM      Profile for Paul Konen   Email Paul Konen   Send New Private Message       Edit/Delete Post 
Here is where you posted the original info.

Anybody attending the Sony 4K Demo

Moderator, a little help needed here [Smile]

 |  IP: Logged

John Pytlak
Film God

Posts: 9987
From: Rochester, NY 14650-1922
Registered: Jan 2000


 - posted 01-20-2005 12:11 PM      Profile for John Pytlak   Author's Homepage   Email John Pytlak   Send New Private Message       Edit/Delete Post 
quote: Lyle Romer
Is it then safe to say that with proper contrast and color, 4K digital projection should eclipse the quality of 35mm film projection.

"Proper contrast and color" have been issues with Digital Cinema. And although a 4K scan is certainly better than a 2K scan, work published by Kodak image scientist Dr. Roger Morton and his team has shown that even higher resolution scanning than 4K is needed to completely capture the detail contained on a 35mm motion-picture negative film without aliasing:

http://www.electronicipc.com/journalez/detail.cfm?code=45390011120508

quote:
An Introduction to Aliasing and Sharpening in Digital Motion Picture Systems
By Roger R. A. Morton, Michelle A. Maurer, and Christopher L. DuMont

May 2003 SMPTE Journal



Abstract
Many interacting factors affect image quality. This paper discusses the measurement of some crucial factors and reviews their interactions. Specifically, how pixel count can interact with the image-quality factors of limiting resolution, aliasing ratio, and shape of frequency response. A new tool designed for onscreen image-quality measurements for both film and electronic projectors is described, and actual examples of aliasing artifacts are shown in still images from digital motion picture systems. It is then illustrated how aliasing artifacts can arise, although Nyquist sampling requirements are satisfied. Finally, some of the interactions between limiting resolution and frequency response shape for still and moving images are explained.



http://www.electronicipc.com/journalez/detail.cfm?code=45390011120705

quote:
Relationships between Pixel Count, Aliasing, and Limiting Resolution in Digital Motion Picture Systems
By Roger R. A. Morton, Christopher L. DuMont, and Michelle A. Maurer

July 2003 SMPTE Journal


Abstract
This paper analyzes how pixel count affects one type of aliasing artifact and image rendition near limiting resolution. A previous paper1 showed that aliasing artifacts take many forms. This paper focuses on the aliasing artifact identified in the ISO 12233 standard and identified as Type A aliasing in the prior paper. A relationship termed “Type A aliasing equation” is presented, which predicts aliasing as measured by the ISO 12233 standard. It is then demonstrated that this equation predicts the best unreconstructed aliasing performance for digital motion picture systems and subsystems, thereby defining one characteristic of an ideal system. The predicted result is then compared by the Type A aliasing equation and the measured aliasing performance of 20 different digital motion picture systems and subsystems. It is also shown that the Type A aliasing equation is superior to the classical Nyquist theory as a predictor of aliasing performance of digital motion picture systems. Finally, the equation is used to compute the minimum number of pixels required for a given aliasing level and limiting resolution, and data is presented to determine the pixel count required to render a given limiting resolution.




 |  IP: Logged

Frank Angel
Film God

Posts: 5116
From: Brooklyn NY USA
Registered: Dec 1999


 - posted 01-20-2005 12:38 PM      Profile for Frank Angel   Author's Homepage   Email Frank Angel   Send New Private Message       Edit/Delete Post 
Well, since we are revisiting, here's a question that I asked but didn't get answered, so I'll give it another shot:
quote:
Ok, I am a bit clueless when it comes to the multiple formats of video in relationship to projection, so this may sound rudimentary, but can someone explain what is meant by 2K and 4K? 4K what? -- pixels, salamis? Is it referring to the nano-mirror module....the video signal....the file size for each frame? And again, whatever it is, why isn't it part of a full spec statement like 1KHz for audio or 256Mb for RAM, or 8MegaPixels etc., as most other specs are stated? And how does the 1K, 2K, 4K relate to HDTV or home video projection equipment? How come you don't see that "K" measurement in the myriad of home video projectors and other equipment, like DVDs? Like, "This InFocus DLP projector has a resolution of .01K."

Inquiring minds want to know.

 |  IP: Logged

John Pytlak
Film God

Posts: 9987
From: Rochester, NY 14650-1922
Registered: Jan 2000


 - posted 01-20-2005 01:00 PM      Profile for John Pytlak   Author's Homepage   Email John Pytlak   Send New Private Message       Edit/Delete Post 
Nominally, it's the number of pixels across the width of the image.

Here's what's available at Kodak Cinesite:

Kodak Cinesite Scanning Submission Form

Remember, with fixed array pixels, the image resolution will usually be less than half the number of pixels, because you need to avoid aliasing artifacts (Nyquist criterion), and you often subsample color information.

 |  IP: Logged

Dominic Case
Expert Film Handler

Posts: 131
From: Sydney NSW Australia
Registered: Aug 2003


 - posted 01-20-2005 05:40 PM      Profile for Dominic Case   Author's Homepage     Send New Private Message       Edit/Delete Post 
quote:
And how does the 1K, 2K, 4K relate to HDTV or home video projection
Usually you find slightly different applications will choose entirely different ways to specify things.

2K and 4k terms are used mainly in the digital inermediate business to indicate the number of pixels across the width of a scanned frame. 4k being 4096; but that is usually the full width of the negative perf to perf, so image width uses less than that. Since there are a number of different aspect ratios in film, it's left at the one common factor - the width.

Still cameras are specified by the total number of pixels, because (a) it's a fixed aspect ratio and (b) the numbers are bigger, and it's a consumer-driven market and (c) the cameras use an area array, whereas most film scanners use a line array (i.e scanning one line at a time, whereas the still camera grabs the whole frame). So a 6 Megapixel still camera is approximately 2840 pixels by 2130 pixels. 6meg is easier for the shop assistant to say.

 |  IP: Logged

Lyle Romer
Phenomenal Film Handler

Posts: 1331
From: Davie, FL, USA
Registered: May 2002


 - posted 01-20-2005 06:10 PM      Profile for Lyle Romer   Email Lyle Romer   Send New Private Message       Edit/Delete Post 
quote: John Pytlak
Remember, with fixed array pixels, the image resolution will usually be less than half the number of pixels, because you need to avoid aliasing artifacts (Nyquist criterion), and you often subsample color information.

John,

Is there a good resource that you can point to which better explains the relationship of Nyquist to fixed pixel arrays. I'm an Electrical Engineer but have no experience with imaging. My understanding of Nyquist has always been with respect to sampling waveforms to avoid aliasing and how if you don't sample at a rate of 2x the frequency, the frequency domain representation of the waveform will "overlap." I'm having trouble understanding how an array of pixels is affected by Nyquist. I'm probably missing something conceptually here but isn't each pixel just a spot represented by a light wave at that point? If I have a tic tac toe board with each box colored in why do I need to scan it at 81 pixels in order to be able to display those original 9 pixels without aliasing?

Thanks to anybody that chimes in here and enhances my understanding.

 |  IP: Logged

John Pytlak
Film God

Posts: 9987
From: Rochester, NY 14650-1922
Registered: Jan 2000


 - posted 01-21-2005 07:25 AM      Profile for John Pytlak   Author's Homepage   Email John Pytlak   Send New Private Message       Edit/Delete Post 
quote: Lyle Romer
My understanding of Nyquist has always been with respect to sampling waveforms to avoid aliasing and how if you don't sample at a rate of 2x the frequency, the frequency domain representation of the waveform will "overlap." I'm having trouble understanding how an array of pixels is affected by Nyquist.
It's exactly the same situation. Whether it's audio (time domain) or an image (spatial domain), when you digitally sample, you need to do it at a frequency at least twice the highest spatial frequency in the image to avoid any risk of "overlap" or aliasing. A fixed pixel array is "sampling" the image, and has the same limitations.

http://www.edn.com/article/CA90388.html?1=1

Nyquist, Harold, "Certain topics in telegraph transmission theory," Transactions of the AIEE #47, February 1928, pg 617 to 644.

quote:
Physicist and engineer Harry Nyquist's 1928 paper on telegraph-transmission theory revealed that complete reconstruction of an N-element signal is possible if you know N/2 sinusoidal components (Reference 1). This theory developed into Nyquist's sampling theorem, which states that complete reconstruction of a waveform is possible from samples taken at a rate greater than twice the highest frequency-harmonic component. If you sample the signal more slowly, an alias results, and information is lost.

The alias is unknown to novice engineers, feared by intermediate-level engineers, and used by expert engineers. Certain conditions are necessary to use these advanced measurement techniques, and simple test methods can reveal otherwise immeasurable time-domain characteristics.


Some references about digital imaging:

Poynton Notes

Poynton Book

Symes Book

Reference Links

Stefan Winkler

Digital Imaging

Kodak Digital Learning Center

Since the "pixels" in film (silver halide grains) are random in their size and spatial distribution, aliasing artifacts are not an issue, unless the film is digitally scanned.

 |  IP: Logged

Frank Angel
Film God

Posts: 5116
From: Brooklyn NY USA
Registered: Dec 1999


 - posted 01-21-2005 11:49 AM      Profile for Frank Angel   Author's Homepage   Email Frank Angel   Send New Private Message       Edit/Delete Post 
As always, thanks for the info guys. And someday John is going to tell us exactly what vitamins he takes soas to allow him to retain an entire reference library in his brain!

Now all someone has to do is figure out a good way to count the grains across a frame of film so as to match apples and apples. 2K, 4K, 8M video.....168K film....something like that. [Wink]

 |  IP: Logged

Mattias Ohlson
Expert Film Handler

Posts: 180
From: Falun, Sweden
Registered: Mar 2003


 - posted 01-21-2005 04:11 PM      Profile for Mattias Ohlson   Email Mattias Ohlson   Send New Private Message       Edit/Delete Post 
The logical thing would be so scan at 8k and store that as a 4k DI master file. This is possibly what will happen. No need to store for a long time at 8k and the 4k file gets smoothed.

 |  IP: Logged

Lyle Romer
Phenomenal Film Handler

Posts: 1331
From: Davie, FL, USA
Registered: May 2002


 - posted 01-22-2005 10:26 AM      Profile for Lyle Romer   Email Lyle Romer   Send New Private Message       Edit/Delete Post 
Thanks for the information and links John. I don't know why I never really thought of digital pictures as a "sampling system" in the same way as digital audio. Now I get it. Thanks for the links, now I can buy some cool books and have my girlfriend think I'm even more of a geek than she already does.

After reading through some of the links, I've determined that there is no quantifiable answer to the question that is asked very often. That is, what resolution is equal to film?

It appears the question needs to be, what resolution of d-cinema will look as good or better than film to a person viewing it from a normal viewing distance? It looks like the only way to find out is to subjectively test many people with each d-cinema system. I guess it is more than just the resolution, it's a combination of resolution, image processing and compression. I guess even if the d-cinema resolution is not capturing all of the negative detail, if that detail could only be seen in a still frame from .002 screen heights, it doesn't really matter.

I was always one that wanted to quantify a particular resolution for the holy grail of d-cinema. I guess the only way to find out is to wait and see each higher resolution. Maybe 4k d-cinema will look as good as 70 mm film even though it won't have anywhere near the detail.

 |  IP: Logged

Mattias Ohlson
Expert Film Handler

Posts: 180
From: Falun, Sweden
Registered: Mar 2003


 - posted 01-22-2005 11:22 AM      Profile for Mattias Ohlson   Email Mattias Ohlson   Send New Private Message       Edit/Delete Post 
I have read that at the highest detail level lenses in aquisition and projecton is a factor that affects the real detail shown in the end.
The same person has said that the type of shutter used in filming and the speed it works at is very important to get alot of detail, think about the camera that is not perfectly still and objects that are moving relative to the camera.

 |  IP: Logged

John Pytlak
Film God

Posts: 9987
From: Rochester, NY 14650-1922
Registered: Jan 2000


 - posted 01-24-2005 04:46 PM      Profile for John Pytlak   Author's Homepage   Email John Pytlak   Send New Private Message       Edit/Delete Post 
quote: Mattias Ohlson
I have read that at the highest detail level lenses in aquisition and projecton is a factor that affects the real detail shown in the end.
The same person has said that the type of shutter used in filming and the speed it works at is very important to get alot of detail, think about the camera that is not perfectly still and objects that are moving relative to the camera.

Modern camera and projection lenses come pretty close to being diffraction limited, which is about as sharp as you can get. Usually there is significant response out to well beyond 100 cycles per millimeter at the f/stops normally used for cinematography and projection.

The "motion blur" of film shot at 24fps with a 180-degree shutter (1/48 second exposure time) is normally considered a good thing, and part of the desired "film look". Using a smaller shutter angle is often used to produce a disconcerting effect, of fast moving objects being too sharp and "stroby". Motion blur occurs only when the camera or scene is moving --- stationary objects have full sharpness without blur. Human vision EXPECTS moving images to have some motion blur (due to persistence of vision).

 |  IP: Logged

Dominic Case
Expert Film Handler

Posts: 131
From: Sydney NSW Australia
Registered: Aug 2003


 - posted 01-24-2005 08:41 PM      Profile for Dominic Case   Author's Homepage     Send New Private Message       Edit/Delete Post 
quote: Lyle Romer
After reading through some of the links, I've determined that there is no quantifiable answer to the question that is asked very often. That is, what resolution is equal to film?

Hallelujah! And from an engineer, too [Cool] Thank you! Administrator, please pin this message to the top of the topic for ever. It is exactly the truth. There is no simple objective answer to the question, and yet people continue to insist that "film is 4K" or whatever.

quote: Lyle Romer
If I have a tic tac toe board with each box colored in why do I need to scan it at 81 pixels in order to be able to display those original 9 pixels without aliasing?

I guess JP's references have answered that for you - but for anyone who hasn't got through all of that yet, the answer is relatively simple. If you line up the tic-tac-toe box exactly with the pixels, then it's true, you only need 3x3=9 pixels to represent it perfectly. (in other words, one pixel per square). If you move it half a pixel to the right and half a pixel up, then every pixel will include a quarter of each of four squares in the box. If they are coloured in a checkerboard pattern, then each pixel will show the same shade of 50% grey. The digital image of the pattern would thus be totally invisible. You need twice the number of pixels in each direction to be sure that the image won't average itself out to nothing.

A film image, with randomly-placed grains, might capture some of the squares in one frame, others in the next, and so on: even if the smallest grains were the same size as the squares in the tic-tac-toe subject, you'd still see the detail of the image over a few frames. But as film has small grains as well as large ones, you are even better off.

 |  IP: Logged

Mattias Ohlson
Expert Film Handler

Posts: 180
From: Falun, Sweden
Registered: Mar 2003


 - posted 01-27-2005 08:33 AM      Profile for Mattias Ohlson   Email Mattias Ohlson   Send New Private Message       Edit/Delete Post 
Do you expect TI will announce a 4k projections system together with one of its partners at Showest in March. This in anticipation of Star Wars III in May.

What is most likely
1 one 4k sxrd projector will be used to project Star Wars III to paying customers this year
or
2 a 4k TI solution will be used to play Star Wars III to paying customers this year

 |  IP: Logged



All times are Central (GMT -6:00)
This topic comprises 3 pages: 1  2  3 
 
Post New Topic  Post A Reply Close Topic    Move Topic    Delete Topic    next oldest topic   next newest topic
 - Printer-friendly view of this topic
Hop To:



Powered by Infopop Corporation
UBB.classicTM 6.3.1.2

The Film-Tech Forums are designed for various members related to the cinema industry to express their opinions, viewpoints and testimonials on various products, services and events based upon speculation, personal knowledge and factual information through use, therefore all views represented here allow no liability upon the publishers of this web site and the owners of said views assume no liability for any ill will resulting from these postings. The posts made here are for educational as well as entertainment purposes and as such anyone viewing this portion of the website must accept these views as statements of the author of that opinion and agrees to release the authors from any and all liability.

© 1999-2018 Film-Tech Cinema Systems, LLC. All rights reserved.