| |
 |
|
|
Science Forum Index » Compression Forum » decode Truncated binary encoding
Page 1 of 1
|
| Author |
Message |
| LiloLilo |
Posted: Sat Apr 26, 2008 11:51 am |
|
|
|
Guest
|
|
| Back to top |
|
| Stefano Brocchi |
Posted: Sat Apr 26, 2008 11:59 pm |
|
|
|
Guest
|
On Apr 26, 6:51 pm, "LiloLilo" <danilobrambi...@tiscali.it> wrote:
Hello,
This shouldn't be possible: if for the decoder all the possible codes
up to the power of two can be used, then there is no way for him to
distinguish a code with less bits from a prefix of a longer code.
An example: on the last example of the wiki page, 0 can be encoded
with two bits because we know there are no codes with prefix 00, as
the encoder knows there are 7 symbols in the alphabet. If the decoder
does not know n, then for him 00 could just be a prefix on a three-bit
symbol.
Greetings
Stefano |
|
|
| Back to top |
|
| LiloLilo |
Posted: Sun Apr 27, 2008 6:59 am |
|
|
|
Guest
|
OK. This makes Truncated binary encoding userful only for quite long
sequences of words, because then I need to add one more word at the
beginning of the transmission with value of n, or almost the value of u =
2^k - 3 which is always 1 bit shorter then n.
For example, if I have a very short sequence of words, say 4 words, with an
alpabeth of n = 10, using standard binary I would need 16 bits. Using
Truncated binary encoding I would need 3 bits to transmit u, then from 3 to
4 bits to transmit each of the 4 words, and this makes quite always a longer
stream.
"Stefano Brocchi" <stefano.brocchi@researchandtechnology.net> wrote in
message
news:4458a1ca-6ec3-42be-98ed-2d43cd1c5ae6@34g2000hsf.googlegroups.com...
Quote: On Apr 26, 6:51 pm, "LiloLilo" <danilobrambi...@tiscali.it> wrote:
Hi,
is the Truncated binary
encodinghttp://en.wikipedia.org/wiki/Truncated_binary_encodingdecodable
when the
receiver doesn't know the size of the alphabet n, but only knows the next
power of 2 of n?
Thank you
Hello,
This shouldn't be possible: if for the decoder all the possible codes
up to the power of two can be used, then there is no way for him to
distinguish a code with less bits from a prefix of a longer code.
An example: on the last example of the wiki page, 0 can be encoded
with two bits because we know there are no codes with prefix 00, as
the encoder knows there are 7 symbols in the alphabet. If the decoder
does not know n, then for him 00 could just be a prefix on a three-bit
symbol.
Greetings
Stefano |
|
|
| Back to top |
|
| Stefano Brocchi |
Posted: Sun Apr 27, 2008 11:41 pm |
|
|
|
Guest
|
On Apr 27, 1:59 pm, "LiloLilo" <danilobrambi...@tiscali.it> wrote:
Quote: OK. This makes Truncated binary encoding userful only for quite long
sequences of words, because then I need to add one more word at the
beginning of the transmission with value of n, or almost the value of u =
2^k - 3 which is always 1 bit shorter then n.
For example, if I have a very short sequence of words, say 4 words, with an
alpabeth of n = 10, using standard binary I would need 16 bits. Using
Truncated binary encoding I would need 3 bits to transmit u, then from 3 to
4 bits to transmit each of the 4 words, and this makes quite always a longer
stream.
Yes, that's right. Another case where it could be useful is where
limitations on the values can be deduced by the context, as in the
case of a series of decreasing integers.
So long,
Stefano |
|
|
| Back to top |
|
| |
|
Page 1 of 1
All times are GMT - 5 Hours
The time now is Sat Jul 26, 2008 11:30 pm
|
|