|
| |
Compression:
So You Want To Compress:
What kind of Compression system do you have, SpaceSaver, Double Space, DriveSpace,
DriveSpace 3, etc., the format makes a difference or the program you might have. I cann't
even dream of anyone using any program other than Windows-95 Plus ---of which is
DriveSpace 3. We're going to talk about DriveSpace 3 only. If the subject confuses you ---
mail-me. Remember also, I recommended that a Pentium be used because things will get too
slow.
What Is Compression Anyway:
DriveSpace 3 achievies a higher ratio by compressing data (CVF) files in 32k clusters
(instead of 8k) as Dblspace & DrvSpace did. DriveSpace 3 can also store more data
via a improved fragmentation handling. As with earlier CVF versions, a DriveSpace 3
CVF has a name in the form DRVSPACE.nnn, where nnn is the CVF sequence
number in the range 000 through 254.
Using A Floppy Disc:
Remember If you send your floppy to a friend, S/He must have DriveSpace also. So,
go ahead and compress your disc--now you have more space on it-ok. You can use it
just as you did before compressing, In short, DriveSpace mounts the floppy as if it was
un-compressed.
Explain Why The Clean-Up:
Don't compress all your disc/error's, duplicate/files, redundant/files, too many fonts,
screen-savers, wallpaper and etc. I'm going to assume that you know how to use the
FIND command to check for all those misc. files that you need to delete. Remember,
when your getting full, you have to start thinking about all those fat files and kicking
heinie for survival. The DELETE Page and Again, if you don't
know how---mail-me.
Preparing The H/D For Compression:
You need to run ScanDisk (thorough) and then a standard ScanDisk. You might have
to back and forth a couple of times to get the (thorough) finished, but that's ok.
Making Decision's About The Menu:
Go to Programs/Accessories/System Tools/DriveSpace 3 and view the menu. Don't
worry about the advanced tab or adjusting free space, because the drive isn't
compressed yet. Highlight and click on compress and you'll see what the compression
tool can do, it will create another drive and show you the expected MB savings.
Accept it's choice for the time being, because you can adjust free space later. If your
running a 150MHz Pentium or higher, then you can think about checking the button
for ultra and hipack ratio's, other wise select normal compression. The reason being,
everything is going to be a bit slower. This will take awhile, so go ahead if you want
compression.
The M/S Plus System Agent:
The system agent is running all the time in the background compressing files
whenever you save that file. You need to do a little scheduling for the Utilities in
agent's to work, this is totally up to you. I run scandisk everyday after closing all
apps. and defrag every week or so and you should throw in compression once a month
at least and remember that the system/agent is helping out in the background all the
time. Try running your programs for awhile to see if your happy with the speed and
remember it's going to be slower unless your running a high end processor and/or a
dual or quad cpu motherboard.
Adjusting Free Space and Ratio's:
The only thing here is give you an example of a friend's 486/100MHz 1 gig drive PC
and he's done alot of changing back and forth: he was happy with the host/drive
being set at 290-300 MB's. You can experiment for your system or e-mail me.
Back to Top
History of Compression..If you like to know more about it.
Two kinds of lossless compression techniques are Huffman coding, and arithematic coding
are very
common forms.
The research of Ziv and Lempel led to a family of compression methods based on the idea of
replacing a group of characters by a reference to an earlier occurrence of that same
group.
The authors explore one particular form of this compression method which restricts the
references so that they can only point into a finite sliding window of text. Their aim is
clearly to develop a practical compression algorithm. In particular, the authors attempt
to achieve good compression performance, but at the same time they want fast compression
and decompression algorithms with reasonable storage requirements. As far as compression
performance is concerned, the algorithms are definitely successful; the figures reported
in the paper indicate that they achieve compression similar to or better than competing
methods. The compression performance of these algorithms is also good on relatively small
files, whereas many other methods achieve their best performance only on very large files.
Unfortunately, the authors do not provide all the data needed to compare execution time
requirements. I was particularly disappointed to find that they make no speed comparison
with the ZLW (Ziv-Lempel-Welch) method, which is provided on most UNIX systems as the
compress command and is in widespread use.
Maybe I should clarify my heading question...
There are two types of compression, lossy and lossless.
Lossy compression implies that not all of the data is restored upon decompression, but the
signal is "good enough" for human beings not to perceive the difference. A
transform or filter bank and
psychoacoustic model are usally involved. In the case of Twin VQ, from NTT, vector
quantization is also used.
Examples of lossly audio compression are
AC-3
MPEGx of all flavors
PAC
ATRAC
Twin VQ
ADPCM
Then there is lossless compression. This is where the entropy of the data from Shannon's
theorem is
used...huffman coding and arthimetic coding are common techniques used in lossless.
Lossless
compression geared towards audio also manipulates the periodicity of the signal. So, no
"bit" of audio is lost...just how one "reorders" the information, per
bit, is manipulated.
Examples are:
Shorten
HEAR
OggSquish
My question really is asking, "does anyone know of any lossless audio compression
algorithms beyond the three listed above and the companys' that have them?"
Chris Boozer wrote:
lossless compression is no differant than data grade compression for a computer the
algorithim measures the data stream and throws out data based on a formula this formula
when decoded restores all of the extracted data so you save space for storage and lose
nothing however on a computer a backup is paramount with compression since and corruption
destroyes data.on dvd or cd with decode hardware you can store the data and reconstruct it
without loss however the
compression rate is far lower than in implemented dts decoder chips or dolby digital so
the best
compression i've read about is 50% reduction maximum.with blue laser light this is
practicle for audio but not video.
Back to Top
Still want to know more--well here's a Very Good Algorithm Program:
NOTES ON THE LZRW3-A ALGORITHM
==============================
Author : Ross Williams (ross@spam.ua.oz.au).
Date : 15-Jul-1991.
Abstract
--------
This file announces the release of my LZRW3-A algorithm which was
invented on 31-Dec-1990 along with LZRW3. The LZRW3-A algorithm has
the following features:
1 Requires only 16K of memory (for both compression and decompression).
2 The compressor runs about two times faster than Unix compress's.
3 The decompressor runs about three times faster than Unix compress's.
4 Yields a few percent better compression than Unix compress for most files.
5 Allows you to dial up extra compression at a speed cost in the compressor. The speed of
the decompressor is not affected.
6 Algorithm is deterministic.
7 Algorithm is free of patent problems. The algorithm has not been patented (nor will it
be) and is of the LZ77 class which is fairly clear of patents. 8 This implementation in C
is in the public domain.
(Timing tests for the speed comparison were performed on a Pyramid 9820.)
LZRW3-A dominates Unix compress in memory and speed. LZRW3-A dominates
Unix compress in compression for object files and smallish (e.g. 50K) text files, but
yields worse
compression for large complex English text files.
The exact figures for the Calgary corpus on C implementations on my Macintosh-SE are
(percentage remaining, compression speed, decompression speed, memory required during
compression and decompression):
PerRem ComK/S DecK/S ComMem DecMem
LZRW1-A 55.1 % 17 K/s 69 K/s 16 K 0 K
LZRW2 51.5 % 18 K/s 54 K/s 24 K 16 K
LZRW3 50.0 % 20 K/s 33 K/s 16 K 16 K
LZRW3-A 45.8 % 8 K/s 31 K/s 16 K 16 K
I would like to hear from anyone who knows of an algorithm that performs similarly or
better to this one when implemented in C and compiled and run on the same machine on the
same files.
Availability
------------
The only implementation available is in C. It can be found in the following archive in
about mid August 1991.
FTP Archive Access:
Machine : sirius.itd.adelaide.edu.au [IP=129.127.40.3]
Directory : ~pub/compression
Files : lzrw3-a.txt - This file.
lzrw3-a.c - An implementation in C.
Motivation for LZRW3-A
----------------------
LZRW3-A was designed as a direct competitor to Unix compress.
LZRW3-A is identical to the LZRW3 algorithm except that it uses a "deep" hash
table (of depth 8 by default). See the explanation in the comments in the program code for
more details.
Benchmark
---------
Here are the results of applying LZRW3-A.C compiled under THINK C 4.0 and running on a
Mac-SE (8MHz 68000) to the standard Calgary corpus.
+----------------------------------------------------------------+
| DATA COMPRESSION TEST |
| ===================== |
| Time of run : Mon 15-Jul-1991 05:29PM |
| Timing accuracy : One part in 100 |
| Context length : 262144 bytes (= 256.0000K) |
| Test suite : Calgary Corpus Suite |
| Files in suite : 14 |
| Algorithm : LZRW3-A |
| Note: All averages are calculated from the un-rounded values. |
+----------------------------------------------------------------+
| File Name Length CxB ComLen %Remn Bits Com K/s Dec K/s |
| ---------- ------ --- ------ ----- ---- ------- ------- |
| rpus:Bib.D 111261 1 49044 44.1 3.53 8.47 31.19 |
| us:Book1.D 768771 3 420464 54.7 4.38 7.27 30.07 |
| us:Book2.D 610856 3 277955 45.5 3.64 8.51 33.40 |
| rpus:Geo.D 102400 1 84218 82.2 6.58 4.23 15.04 |
| pus:News.D 377109 2 192880 51.1 4.09 7.08 25.89 |
| pus:Obj1.D 21504 1 12651 58.8 4.71 5.23 17.44 |
| pus:Obj2.D 246814 1 108044 43.8 3.50 8.01 28.11 |
| s:Paper1.D 53161 1 24526 46.1 3.69 8.11 30.24 |
| s:Paper2.D 82199 1 39483 48.0 3.84 8.11 32.04 |
| rpus:Pic.D 513216 2 111622 21.7 1.74 10.64 49.31 |
| us:Progc.D 39611 1 17923 45.2 3.62 8.06 29.01 |
| us:Progl.D 71646 1 24362 34.0 2.72 10.74 39.51 |
| us:Progp.D 49379 1 16805 34.0 2.72 10.64 37.58 |
| us:Trans.D 93695 1 30296 32.3 2.59 11.02 38.06 |
+----------------------------------------------------------------+
| Average 224401 1 100733 45.8 3.67 8.29 31.21 |
+----------------------------------------------------------------+
If you have to--ya have to
Back to Index
|