🚀 go-pugleaf

I was wondering the other day how smart the Apple-provided QuickTime compressors
are in doing their frame-differencing compression. Would they do motion
detection?

To investigate this, I created a very simple movie in MacroMind Director 3.0.
This was a black ball, 116 pixels in diameter, moving across a white stage
(342 * 512 pixels in size) from left to right. Just for fun, I drew a
graduated grey-scale highlight with Canvas and stuck it on the ball.

I then exported the animation as a PICS file (Director 3.0 doesn't have
direct QuickTime support) and tried compressing it as a QuickTime movie in
various ways. The animation had I think a total of 25 frames, and the PICS
file size was 192K (as opposed to 4K or something equally ridiculous for the
original Director file). I had enabled Director's frame-differencing facility,
where each frame of the PICS file after the first only redrew the part of the
previous picture that was different.

I put this data through the Animation compressor, which works by run-length
encoding. At 8 bits per pixel, the resulting QuickTime movie was 196K in size.
Allowing for some extra overhead for the QuickTime movie format, I'd say
the Animation compressor had done about as good a job as Director had of
picking up the differences between frames.

The other interesting thing was that I'd set the spatial and temporal
compression quality settings down to their absolute minimum. The Animation
compressor is supposed to have a lossy mode, but I think this does not apply
to pixel depths of 8 bits or less.

I then tried the Graphics compressor, which only supports 8-bit-per-pixel
images, but apparently works well with dithered images. With the quality
settings on minimum, I was able to get the size of the movie file down to
52K, without any really noticeable quality loss.

Finally I went back to the Animation compressor, and specified a depth of
16 bits per pixel. This means the Image Compression Manager had to depth-convert
the frames of the animation up from the original 8 bits per pixel before
compressing them. I was surprised to find that the size of the movie file
was now 119K! I'd say it was definitely getting into lossy compression now.
When I viewed the movie, I saw a bit of graininess in the highlight on the
ball, but that was probably just an artifact of the "fast dithering" mode
for displaying the movie on my 8-bit screen.

By the way, I also discovered why Director is so slow at exporting its
animations to PICS files: it closes and reopens the file after adding each
frame! What a dumb thing to do, in what is otherwise such a neat program.
In any case, I've sent the order off for the 3.1 upgrade--I'm looking forward
to seeing what its direct QuickTime support is like.

Lawrence D'Oliveiro                       fone: +64-7-856-2889
Computer Services Dept                     fax: +64-7-838-4066
University of Waikato            electric mail: ldo@waikato.ac.nz
Hamilton, New Zealand    37^ 47' 26" S, 175^ 19' 7" E, GMT+12:00