| Science Forum Index » Compression Forum » new algo de compression... |
|
Page 1 of 1 |
|
| Author |
Message |
| remy... |
 Posted: Fri Oct 09, 2009 2:17 am |
|
|
|
Guest
|
|
| Back to top |
|
|
|
| Matthias Gallé... |
| Posted: Fri Oct 09, 2009 7:09 am |
|
|
|
Guest
|
Bonjour Remy:
remy wrote:
[quote:f33e22b6f0]hello
yes/no ?
Let me see if I understood right: you have a window for your sequence[/quote:f33e22b6f0]
and you look if the symbols in this window can be expressed with less
then 8 bits. If this is the case you encode it with the minimal amount,
and you move the window.
Is this correct?
Have some results over standard benchmark?
[quote:f33e22b6f0]
http://remyaumeunier.chez-alice.fr/compression_En.php
I speak very little English
thank you remy
[/quote:f33e22b6f0]
--
Matthias Gallé
Project Symbiose
Centre de Recherche INRIA Rennes - Bretagne Atlantique,
Campus de Beaulieu, 35042 Rennes cedex, France
tel: (33|0) 2 9984 7523
http://www.irisa.fr/symbiose/matthias_galle |
|
|
| Back to top |
|
|
|
| Thomas Richter... |
| Posted: Fri Oct 09, 2009 7:27 am |
|
|
|
Guest
|
Matthias Gallé schrieb:
[quote:27116087c4]Bonjour Remy:
remy wrote:
hello
yes/no ?
Let me see if I understood right: you have a window for your sequence
and you look if the symbols in this window can be expressed with less
then 8 bits. If this is the case you encode it with the minimal amount,
and you move the window.
Is this correct?
Have some results over standard benchmark?
[/quote:27116087c4]
For a related work: Look at the "tagtree" encoding in JPEG 2000.
So long,
Thomas |
|
|
| Back to top |
|
|
|
| remy... |
| Posted: Fri Oct 09, 2009 7:48 am |
|
|
|
Guest
|
Matthias Gallé a écrit :
[quote:b088d1e774]Bonjour Remy:
remy wrote:
hello
yes/no ?
Let me see if I understood right: you have a window for your sequence
and you look if the symbols in this window can be expressed with less
then 8 bits. If this is the case you encode it with the minimal amount,
and you move the window.
[/quote:b088d1e774]
[quote:b088d1e774]Is this correct?
[/quote:b088d1e774]
yes
[quote:b088d1e774]Have some results over standard benchmark?
http://remyaumeunier.chez-alice.fr/codeSourceGene/Compression.zip[/quote:b088d1e774]
ls -l lena*
-rw------- 1 remy remy 786572 2009-01-16 16:23 lena_std.tif
-rw-r--r-- 1 remy remy 777629 2009-02-23 15:03 lena_std.tif.ar
remy |
|
|
| Back to top |
|
|
|
| remy... |
| Posted: Fri Oct 09, 2009 7:49 am |
|
|
|
Guest
|
Thomas Richter a écrit :
[quote:cbf673939e]Matthias Gallé schrieb:
Bonjour Remy:
remy wrote:
hello
yes/no ?
Let me see if I understood right: you have a window for your sequence
and you look if the symbols in this window can be expressed with less
then 8 bits. If this is the case you encode it with the minimal amount,
and you move the window.
Is this correct?
Have some results over standard benchmark?
For a related work: Look at the "tagtree" encoding in JPEG 2000.
[/quote:cbf673939e]
I made the lossless compression
remy
--
http://remyaumeunier.chez-alice.fr/ |
|
|
| Back to top |
|
|
|
| Thomas Richter... |
| Posted: Fri Oct 09, 2009 8:02 am |
|
|
|
Guest
|
remy schrieb:
[quote:0ba3d2096a]Thomas Richter a écrit :
Matthias Gallé schrieb:
Bonjour Remy:
remy wrote:
hello
yes/no ?
Let me see if I understood right: you have a window for your sequence
and you look if the symbols in this window can be expressed with less
then 8 bits. If this is the case you encode it with the minimal amount,
and you move the window.
Is this correct?
Have some results over standard benchmark?
For a related work: Look at the "tagtree" encoding in JPEG 2000.
I made the lossless compression
[/quote:0ba3d2096a]
The tag tree compression in JPEG 2000 *is* lossless. First read, then
think. (-;
Greetings,
Thomas |
|
|
| Back to top |
|
|
|
| remy... |
| Posted: Fri Oct 09, 2009 8:18 am |
|
|
|
Guest
|
Thomas Richter a écrit :
[quote:bb2440aaf4]remy schrieb:
Thomas Richter a écrit :
Matthias Gallé schrieb:
Bonjour Remy:
remy wrote:
hello
yes/no ?
Let me see if I understood right: you have a window for your sequence
and you look if the symbols in this window can be expressed with less
then 8 bits. If this is the case you encode it with the minimal amount,
and you move the window.
Is this correct?
Have some results over standard benchmark?
For a related work: Look at the "tagtree" encoding in JPEG 2000.
I made the lossless compression
The tag tree compression in JPEG 2000 *is* lossless. First read, then
think. (-;
[/quote:bb2440aaf4]
je ne suis pas assez bon en anglais, tu ne m'en voudras pas hein
effectivement le jpg 2000 peut éventuellement faire du codage sans
perte, mais je ne fais pas le lien entre le jpg2000 et" mon algo
"sauf si vous me l'expliquez bien sur
google translation
actually jpg 2000 may possibly be the lossless coding, but I do not
link the jpg2000 and "my algo" unless you explain to me of course
remy
--
http://remyaumeunier.chez-alice.fr/ |
|
|
| Back to top |
|
|
|
| Thomas Richter... |
| Posted: Fri Oct 09, 2009 8:29 am |
|
|
|
Guest
|
remy schrieb:
[quote:ba86f29cee]je ne suis pas assez bon en anglais, tu ne m'en voudras pas hein
[/quote:ba86f29cee]
I could offer German, but unfortunately, I do not speak French.
[quote:ba86f29cee]effectivement le jpg 2000 peut éventuellement faire du codage sans
perte, mais je ne fais pas le lien entre le jpg2000 et" mon algo
"sauf si vous me l'expliquez bien sur
google translation
actually jpg 2000 may possibly be the lossless coding, but I do not
link the jpg2000 and "my algo" unless you explain to me of course
[/quote:ba86f29cee]
No - things are different. I'm not saying that you should use JPEG 2000,
I only want to draw your attention to a minor part of the standard which
is a compression scheme of its own.
The tagtree coding is only a minor part of the complete algorithm: It is
used to encode which bit planes of which code blocks appear in the final
bit stream, and it looks close to what you are doing. Again, it does not
compress images. It is used in JPEG 2000 to compress side information.
See Annex B.10 of the standard for details.
So long,
Thomas |
|
|
| Back to top |
|
|
|
| remy... |
| Posted: Fri Oct 09, 2009 8:42 am |
|
|
|
Guest
|
Thomas Richter a écrit :
[quote:5c80db2f8d]remy schrieb:
je ne suis pas assez bon en anglais, tu ne m'en voudras pas hein
I could offer German, but unfortunately, I do not speak French.
effectivement le jpg 2000 peut éventuellement faire du codage sans
perte, mais je ne fais pas le lien entre le jpg2000 et" mon algo
"sauf si vous me l'expliquez bien sur
google translation
actually jpg 2000 may possibly be the lossless coding, but I do not
link the jpg2000 and "my algo" unless you explain to me of course
No - things are different. I'm not saying that you should use JPEG 2000,
I only want to draw your attention to a minor part of the standard which
is a compression scheme of its own.
The tagtree coding is only a minor part of the complete algorithm: It is
used to encode which bit planes of which code blocks appear in the final
bit stream, and it looks close to what you are doing. Again, it does not
compress images. It is used in JPEG 2000 to compress side information.
See Annex B.10 of the standard for details.
[/quote:5c80db2f8d]
thank you for the info and references
remy
--
http://remyaumeunier.chez-alice.fr/ |
|
|
| Back to top |
|
|
|
| remy... |
| Posted: Mon Oct 12, 2009 1:56 am |
|
|
|
Guest
|
Thomas Richter a écrit :
[quote:3ec22104a1]
See Annex B.10 of the standard for details.
[/quote:3ec22104a1]
Are you referring to coding huffman tree?
remy
--
http://remyaumeunier.chez-alice.fr/ |
|
|
| Back to top |
|
|
|
| Thomas Richter... |
| Posted: Mon Oct 12, 2009 5:05 am |
|
|
|
Guest
|
remy schrieb:
[quote:b4aa6bef32]Thomas Richter a écrit :
See Annex B.10 of the standard for details.
Are you referring to coding huffman tree?
[/quote:b4aa6bef32]
No, it's not per se a Huffman coding. It is an efficient encoding of a
two-dimensional matrix containing numbers. The idea is related to
quadtree coding:
Each entry of the matrix is a leaf of a two-dimensional tree where four
neighboring entries (leafs) are combined into one node. The number in
the parent node is the minimum of the numbers in the children. This
procedure is repeated recursively until only one node (the root) is
left. Encoding now proceeds by first coding the parent, and then
transmitting the increment from the parent to the corresponding childs.
Details are found in the standard, as referenced above; the trick is
that only the minimal information required to reconstruct the next
requested entry in the matrix is transmitted.
So long,
Thomas |
|
|
| Back to top |
|
|
|
| remy... |
| Posted: Mon Oct 12, 2009 6:25 am |
|
|
|
Guest
|
Thomas Richter a écrit :
[quote:9a8384ac60]the trick is
that only the minimal information required to reconstruct the next
requested entry in the matrix is transmitted.
[/quote:9a8384ac60]
ok now I understand better the link
Thank you for explanations
we can also see a link between my algo
and rle Run-length encoding
aaaabbbccccc -> 4a3b5c
where there is a factorisation of the "space entropy" = the table
remy
--
http://remyaumeunier.chez-alice.fr/ |
|
|
| Back to top |
|
|
|
|
