Re: A Plea For Film Digitization (was: Digital restoration of silent films)

From: David P. Hayes
Newsgroups: alt.movies.silent
Date: Friday, January 16, 1998 5:02 PM (ChaneyFan) in messag
<> begins with a quote from me:
>>>(It doesn't help that the film grains are not in the same position from one film frame to the next. The manufacturing process can't guarantee that. Digital video gives each pixel an "address" that places it at a screen location that remain fixed from one frame to the next.)

>…Also, David said that a feature film would require a staggering 106 DVDs for
>storage. I say, big deal. Within 5 years the capacity of these will
>quadruple, and write-once DVD units will be available, and at that point it
>will take 27 disks at $20 each ($540) to digitally preserve a feature.

I also didn't think 106 DVDs was a particularly high number--it pales against the 1,068,988 3.5" floppy disks cited in the article to which I was responding. Actually the number of DVD-ROMs equivalent to 1.068 million 3.5"s would actually be lower. I arrived at 106 by dividing by 14GB per DVD. The actual number turns out to be 17GB, a difference of about 17%. My source here is the catalog of Computability, which has a web link for this particular product line (Creative brand PC-DVD-ROM drives (PC-DVD Encore)) at:

A $20 per unit cost for DVD-ROM blanks may be too high. Although they are as yet unavailable, recordable blank CD-ROMs have been being advertised for $3 each (even less with rebates, but I don't count those here because of the quantity limits therein), and given the similar specs, mass-manufacturing of DVD-ROM will keep the price down to insignificantly more than recordable CD-ROM.

Bob Birchard was concerned about storage on large numbers of separate units, such leading to misplacing of some portions of a movie. A solution to this, given the relatively small number of disks, would be storing and playing the movie from a locked container which houses all of the discs for a single movie. This single container would be similar to a small jukebox, audio CD-changer cartridge, or the rectangular trays used to store slide photographs.

>The point is that I can tell you that film emulsions, at the microscopic
>level, are incredibly irregular. A typical silver halide crystal is oddly shaped
>(not at all round…more like a splop of paint) and varies in size from about
>0.5-5 microns depending on the temperature you developed it at. This fits with
>David's assertion that computer digitization may actually be using higher
>resolution than it needs to. I posted the mathematical analysis on the
>amount of storage space needed per frame a few weeks back and am too lazy to
>reproduce it now, but I think it was around 15-20 MB per frame, but the point is that
>you could probably store the information to a resolution indistinguisable by the
>human eye with 20% of that.

Everyone who has posited a statistic on number of MBs has come up with from 12-20MB, so we seem to have a decent ballpark figure. If 0.5 microns is the size of the smallest grain, and an average is 2.5, then figuring 1.8cm x 2.1cm for the image area of a 35mm sound film (this figure comes from a measurement I made within the last minute)(silent films have a slightly larger picture area), we can make computation, which, when combined with knowledge about the bit-size needed for the most accurate color attainable, gives the answer.

(Theoretically, there is no limit to the number of colors and hues possible by forever increasing the number of bits per byte, but in actual practice, we must remember that film itself does not completely accurately reproduce the colors and hues on the source material. There is always a tolerance level. Once we determine what that level is for film, we can apply the same numbers for video. The programmers can always opt for a higher standard for video, but for purposes of cost comparison, it is only fair to determine the financial outlay for the same degree of quality.)

Compression techniques can indeed "store the information to a resolution indistinguishable by the human eye with 20% of that [of the original source]." However, there may be a dangerous precedent set by calculating these storage-savings in from the beginning. Let's make sure we are storing every bit of data we could possibly ever need. In defense of compression techniques, I've viewed some of the images published on the net, and on an SVHS monitor (1024x768; 65,535 colors), these display images that seem to take advantage of every pixel available and yet in .JPEG form take up 300KB.

David Hayes


Return to Table of Contents

Go to next article