Home of Professional Cinematography since 1996

HD Data Capture

Published : 16th October 2003


Rolling 65mm at 270 feet per minute killed Showscan.

Greetings :

By far the best projected 2D images I have ever seen was IMAX-HD(48 fps capture and projection)but it also died for similar but greater cause(try 810 per min). It smoothed out all the motion judder etc. inherent in the large formats.

Ernie McNabb csc DOP/Steriographer shot the tests and one film that was shown at the Canadian Pavilion at the World's Fair in Seville. He was anxious to try this in 3D and I believe actually did some tests. But 810 ft per min of 70 mm film times 2 gets up into the rather exotic cost region and no surprise; no ImaxHD-3D film was ever actually completed.

Alas; Ernie was stricken with cancer and passed on at far too early an age(mid-50s) earlier this year.

And speaking of data and Imax, Gord Harris who used to work there and was a physicist by education and had a dry sense of humour used to give comparisons of the cost of storage of data on various media from floppies, to CD to hard drives(before recent price drops) to storing data on a frame of Imax. And hey wouldn't you know it, a frame Imax film wins hands down at something like 0.03 cents per Mb(don't quote me on this).

Thanks again for the continued sharing of info.

Regards,

Lance Carlson
Cinematographer/Consultant
Toronto



If one makes a distinction between production data and delivery/exhibition data a very compelling case can be made for various lossy compression technologies. I have seen 4K data compressed to 800K files that, when expanded, are nearly identical to the full-res source. To the average viewer these images would be as close to perfection as they could perceive.

Some have made the argument that you could do 4K data, compressed 50 to 1for production/post-production use. The theory being that the real goal is to deliver a 2K product and that, even with lossy compression, the images look great and are easily manipulated. Maybe that's one possible future path. Don't know.

Martin Euredjian
eCinema Systems, Inc.
www.ecinemasys.com



>Some have made the argument that you could do 4K data, compressed >50 to 1for production/post-production use.

I would agree that compressed images, if done properly, will _look_ close to what the uncompressed footage looked like. However, there is a huge difference between using such footage for final delivery ( = end product) or using it as the input to postproduction.

Footage that looks "perfect" to the casual viewer will cause a tremendous amount of problems when used in a compositing or digital color grading pipeline. Basically all compression today massively reduces the amount of color information in the image. This can't be undone by using a higher pixel count. Perception-based compression simply is not suitable for footage that will be processed mathematically.

Anything below 10 bit uncompressed (or losslessly compressed) log is a huge step backwards from the status quo. I personally would like to see the industry go _beyond_ the status quo - this means lossless compression, 16 bit or floating point precision and 4K res for the initial data that goes into post.

For distribution to the cinema, this is a totally different ballgame. Quite frankly in most cinemas outside "spanking-new-release-print-carefully-adjusted-projector" Hollywood even a JPEG-compressed 1K image would be an improvement over what they show today...

I always wonder why they don't simply connect a DVD player to a $10K beamer to revolutionize the cinema experience over here in Germany...

But for post, we need resolution : First and foremost color depth – then pixel count.

Cheers,

Lin Sebastian Kayser IRIDAS
Adalbertstr. 32 D-80799 Munich
Germany
www.iridas.com



Lin Sebastian Kayser wrote :

>I would agree that compressed images, if done properly, will _look_ >close to what the uncompressed footage looked like.

That's precisely why I started my post with : "If one makes a distinction between production data and delivery/exhibition". Maybe it wasn't stated clearly enough.

Once you get to the point where all the images are destined for is to have someone look at them, compression is great. If you have to process the images you could run into problems. Simple cutting, for example, is not necessarily problematic (if you don't go through repeaded expand-compress cycles) but almost anything beyond that could be.

Martin Euredjian
eCinema Systems, Inc.



Martin Eurediian wrote :

>I have seen 4K data compressed to 800K files that, when expanded, are >nearly identical to the full-res source.

I've seen those demos as well, and very impressive they are too.

I'll start to believe in it when Peter shows moving pictures using these techniques.

I think that there are probably artefacts that are irrelevant to a single image but when the irrelevant variations vary 24 times a second they suddenly become very relevant.

Maybe that's why we never see moving images at these demos?

Cheers

Geoff Boyle FBKS
Director of Photography
EU Based
www.cinematography.net



Martin Euredjian writes :

>Some have made the argument that you could do 4K data, compressed >50 to 1for production/post-production use.

That's an interesting idea, because the more pixels in the image, the less visible the artefacts of compression, processing, etc.

It might be interesting to compare 2K compressed 12.5:1 and 4K compressed 50:1. The net amount of data would be the same, but which would look better, and which would better survive processing and tweaking?

Dan Drasin
Producer/DP
Marin County, CA



Lin Sebastian Kayser writes :

>in most cinemas outside "spanking-new-release-print-carefully->adjusted-projector" Hollywood even a JPEG-compressed 1K image >would be an improvement over what they show today...

At the Mill Valley Film Festival last year I saw the indie feature NEW SUIT (superbly DP'd by our very own David Mullen) projected theatrically using a large, portable projector. From about 3/4 of the way back in the house I couldn't tell it wasn't film. Later I found out not only that it had been shot on HD, but that it was being played back in NTSC Digibeta!

Dan Drasin
Producer/DP
Marin County, CA



Geoff Boyle wrote :

>I think that there are probably artefacts that are irrelevant to a single >image but when the irrelevant variations vary 24 times a second they >suddenly become very relevant.

That's pretty much the problem with any technology that does not transport/acquire the full image data set, Bayer filters included. They work great for still cameras but are likely to require very sophisticated processing in order to avoid issues when motion pictures come into play.

Martin Euredjian
eCinema Systems, Inc.



Dan Drasin wrote :

>It might be interesting to compare 2K compressed 12.5:1 and >4Kcompressed 50:1. The net amount of data would be the same, but >which would look better, and which would better survive processing and >tweaking?

It's been done ... on stills, as Geoff said. I'll have to go back to the source (Peter Swinson, ex Cintel) and ask about manipulation results.

Martin Euredjian
eCinema Systems, Inc.



Geoff Boyle said :

> It's been done...on stills

Big difference in temporal resolution and spatial resolution though. Hence why Panasonics camera with better spatial resolution but with what appears to be 'lesser' numbers' actually looks the same if not better. But take a still from a Sony camera and you've got a pretty damn nice still camera, just mighty expensive.

Walter Graff
Producer, Director, Creative Director, Cinematographer
HellGate Pictures, Inc.
www.film-and-video.com



Dan Drasin wrote :

>It might be interesting to compare 2K compressed 12.5:1 and >4Kcompressed 50:1.

A follow-up to this. I went back to the source and got the straight poop. I was wrong on the still-only part. The tests were written back to film as a split screen (full-res on one side and 50 to 1 expanded material on the other side). Prints were struck and the material was projected. About 1500 people have seen the material and I'm told not a single person has been able to point out where the split might be.

I might have an HDCAM or D5 transfer of this film soon for work I'm doing with display processing technology.

Martin Euredjian
eCinema Systems, Inc.