On Aug 09 2000, Krzys Majewski wrote:
> -rw-r--r-- 1 krzys krzys 118700 Jul 31 17:28 hip1302mp3.mp3
> -rw-rw-r-- 1 krzys krzys 1308716 Aug 9 10:05 hip1302mp3.wav
> -rw-rw-r-- 1 krzys krzys 117718 Aug 9 10:06 hip1302mp3.wav.gz
>
> So what's the point of .mp3? -chris
The point is that gzip can't always have such high compression
rates with generic wav files.
There's a (very, should I add) astute of mesuring how
(theoretically) compressable a file is. This measure is called
its entropy (and both the name of the measure and its
motivation come from thermodynamics): files with *higher*
entropy can be *less* compressed and files with *lower*
entropy can be *more* compressed.
(Intuitively, you can regard the entropy of a file like a
degree of disorder or lack of structure).
Since gzip (or bzip2, for that matter) is a lossless
compression program, it won't throw any data from the file to
generate its compressed output. If the file happens (in a very
unlikely case) to have a good deal of structure that gzip
happens to recognize, then it can compress the file a lot. But
most of the time, that isn't the case.
On the other hand, MP3 compressors don't take the structure of
the file as strictly as lossless compressors would do: they
are lossy compressors (which means that the compressed file
doesn't have to have the same contents as the original file).
So, lossy compressors are able to throw small elements right
out the window and still consider the compression process OK.
These small, local contents in the input file, if infrequent
enough, may mess with a lossless compressor's perception of
the structure of a file (making the entropy of the file
possibly high) and make it compress very little (if any amount
at all). The MP3 compressors are made so that the details they
ignore aren't supposed to make much difference to the listener
if they are ignored (this is what is called "perceptual
compression").
In this sense, lossy compressors are concerned with the "big
picture" of a file, without taking into account the smaller
details of a file (which, again, *would* be considered by
lossless compressors).
This explains why MP3 compressors may have poor efficiency
with one file or another, but are a big plus with sound files
in general. But then, there are also other factors to consider
like with which bitrate (i.e., quality) was the MP3 file
generated, how efficient was the encoder used etc.
Data compression is really a fascinating subject, in my
opinion. :-)
[]s, Roger...
--
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
Rogerio Brito - [EMAIL PROTECTED] - http://www.ime.usp.br/~rbrito/
Nectar homepage: http://www.linux.ime.usp.br/~rbrito/nectar/
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
