NetNews Usenet Archive 1992 #31/NN_1992_31.iso/spool/rec/photo/22570

Internet Message Format | 1992-12-26 | 3.2 KB

Path: sparky!uunet!spool.mu.edu!uwm.edu!ogicse!mintaka.lcs.mit.edu!ai-lab!life.ai.mit.edu!bson
From: [e-mail] (Jan Brittenson)
Newsgroups: rec.photo
Subject: Re: What will it really take to replace FILM!
Message-ID: <[e-mail]>
Date: 24 Dec 92 16:21:24 GMT
Article-I.D.: churchy.BSON.92Dec24082124
References: <[e-mail]> <[e-mail]>
Organization: nil
Lines: 57
NNTP-Posting-Host: churchy.gnu.ai.mit.edu
In-reply-to: [e-mail]'s message of 24 Dec 92 00:56:11 GMT

In article <[e-mail]> [e-mail] (Matthew D. Kirkwood) writes:

 > Based on the assumption that a _good_ color film has a resolution of less
 > than 150 lines/mm (this assumption should be very good), the resolution
 > of a 35mm frame is:
 > 36mm*150 lines/mm = 5400 lines
 >        24mm*150 lines/mm = 3600 lines
 >   i.e.  5400 x 3600 pixels on a 35mm slide.   Therefore, a digital camera
 >   should have a minimum of that resolution to beat a really good film.
 >       5400*3600*3 bytes/pixel = 55.6 MBytes
 >   for a 24bit image.

   This is not exactly true, the reason is that film doesn't consist
of square pixels. When talking about film, you shouldn't talk about
resolution, but about resolving power -- same as for lenses.

First:

   Developed film consists of grains, halide ``blobs'', which are
neither round, square, or symmetrical. It is the grain shape and
overlapping patterns that determine the film's characteristic grain
pattern. While it's true that some films can resolve 150 lines/mm, it
doesn't behave like pixels. With film, smaller lines tend to ``melt
together'', with pixels they will still get averaged along with the
surrounding background into one pixel.

   To make digital images look like film, and by that I mean a
_particular_ film, the system will have to outresolve film enough to
adequately copy the effects of the grain structure itself. Remember
that color film has several grain layers that are by no means aligned
with each other.

   Film with a grain structure exhibiting pixel-like (squarish,
non-random structure, etc) properties would be vastly inferior to all
existing film, and would probably be considered complete junk.

Second:

   The resolving power of a digital system depends on the orientation
of the line target. This is because, as I'm sure anyone who isn't
computer-naive knows, pixels are perfectly lined up in rows and
columns -- but film grain isn't. Misalign the target by angles
``atan(prime1/prime2)'', and you can determine the *minimum* resolving
power of the digital recording system.  Notice how it now makes sense
to distinguish between resolution (number of pixels) and resolving
power (number of lines/mm or lp/mm the system can record as distinct
entities). This ought to be readily apparent to any novice programmer
who has ever written line-drawing code using only integer arithmetic.
Remember also that displays and printers don't produce perfectly
separated pixels, so a ``fuss factor'' needs to be added to the
reproduction end of the chain.

The 150MB mentioned per 35mm frame sounds minimal to me.

--
                        -- Jan Brittenson
                           [e-mail]