Home > Archive > Compression > February 2005 > How pegasus losslessly compresses JPEGs by upto 30%
You are viewing an archived Text-only version of the thread.
To view this thread in it's original format and/or if you want to reply to
this thread please [click here]
| Author |
How pegasus losslessly compresses JPEGs by upto 30%
|
|
| rep_movsd 2005-02-16, 8:55 am |
| Hi
Heres a pure surmise on how pegasus shrinks Jpegs.
There are two important clues that come from their blurb,
1) It is pixel identical
2) They want to apply similar stuff to ZIP and other archives
Obviously, the stuff about being able to compress random/ already
compressed data is pure nonsense.
Since they dont lose any data while processing Jpegs, They do not
transform it back to a raw image by dequantizing it. Instead, They
probably undo the Huffman stage, after which they recompress using a
better algorithm. On most files for e.g, PPM based algorithms will do
10s of percent better than order-0 huffman which is used (I presume) in
JPEGS.
As for Zips and others, This is rather like the work I did for a major
Installer company for CAB and PDF files i.e. Decompress and recompress
using better methods.
Anyway, at least they thought of it ( patented it ) first. :(
| |
| Thomas Richter 2005-02-16, 8:55 am |
|
Hi,
> Heres a pure surmise on how pegasus shrinks Jpegs.
There might be several companies called "Pegasus", but
the one that's active in image compression I know
is this one here: http://www.jpg.com
So are we talking about the same folks?
> There are two important clues that come from their blurb,
> 1) It is pixel identical
> 2) They want to apply similar stuff to ZIP and other archives
> Obviously, the stuff about being able to compress random/ already
> compressed data is pure nonsense.
Who's claiming that and where? If that's what they write on any
of the pages mentioned above, and we truely talk about the same
folks, then it's nonsense and I'll try to stop that. In other
words, I would be interested in an URL where you've seen this claim.
> Since they dont lose any data while processing Jpegs, They do not
> transform it back to a raw image by dequantizing it. Instead, They
> probably undo the Huffman stage, after which they recompress using a
> better algorithm. On most files for e.g, PPM based algorithms will do
> 10s of percent better than order-0 huffman which is used (I presume) in
> JPEGS.
There's no big secret about what's going on for the Pegasus JPEG
recompression. It reverses the Huffman coding and applies a better
entropy coder on the data, the ELS coder. The ELS coder is patented,
but IIRC the ELS coder is also available under GPL (unlike the
arithmetic coding, i.e. the QM coder, which is the bonus part here). I
remember you could download this coder from some server, but I forgot
where. I could ask if you care.
> Anyway, at least they thought of it ( patented it ) first. :(
I wonder whether we speak about the same people here, but anyhow...
So long,
Thomas
| |
| rep_movsd 2005-02-16, 8:55 am |
|
Thomas Richter wrote:
> There might be several companies called "Pegasus", but
> the one that's active in image compression I know
> is this one here: http://www.jpg.com
>
> So are we talking about the same folks?
>
Sorry for the confusion
I got it all backwards, The folks i was reffering to are the Stuffit
people.
The claim is there under a thread called 'Lossless compression of
jpegs' on google groups
| |
| SuperFly 2005-02-16, 3:56 pm |
| On 16 Feb 2005 01:05:56 -0800, "rep_movsd" <rep.movsd@gmail.com>
wrote:
-snip-
>Since they dont lose any data while processing Jpegs, They do not
>transform it back to a raw image by dequantizing it. Instead, They
>probably undo the Huffman stage, after which they recompress using a
>better algorithm. On most files for e.g, PPM based algorithms will do
>10s of percent better than order-0 huffman which is used (I presume) in
>JPEGS.
-snip-
Maybe someone can post an uncompressed version of the a10.jpg file
with a short description of the used file structures, just to level
the playing field. So more people might have a go at creating better
data models for it, without first writing a jpeg decoder.
-SF-
| |
| Jim Leonard 2005-02-16, 3:56 pm |
| rep_movsd wrote:
> As for Zips and others, This is rather like the work I did for a
major
> Installer company for CAB and PDF files i.e. Decompress and
recompress
> using better methods.
That company wouldn't happen to be Netopsystems, would it?
| |
| rep_movsd 2005-02-16, 3:56 pm |
|
Jim Leonard wrote:
> rep_movsd wrote:
> major
> recompress
>
> That company wouldn't happen to be Netopsystems, would it?
Lets just say ( wihout naming any names ) that its the first company
which comes to mind when you say installer.
| |
| Jim Leonard 2005-02-16, 3:56 pm |
| rep_movsd wrote:
>
> Lets just say ( wihout naming any names ) that its the first company
> which comes to mind when you say installer.
I will endeavor to Shield my reponse, then.
|
|
|
|
|