| Dag-Erling Smørgrav 2005-12-14, 7:55 am |
| "Lionel Delafosse" <ldelafosse@wanadoo.fr> writes:
> Repeated compression is possible, based on combinations, as I wrote
> it in this group in the past. Let me explain in few words, basic
> idea is to create redundancy...
>
> 1. take a block of 1024 bytes previously compressed.
> 2. divide it in four blocks of 256 bytes.
> 3. sort bytes in each block in ascendant order ; now, each block is
> at least one of the possible reordering combinations from the
> original block ; next, recording what is the corresponding
> combination number
> 4. compress the four blocks with any existing algorithm
> 5. append the 4 combination number with the data from step 4.
> 6. return in 1
>
> When all data has been processed, compress the result again until
> the result satisfy our needs (size, time, etc). Obviously, we have
> to record the number of steps done.
There are two steps here where you add fixed-size data to the output
of the compressor: the four "combination numbers" and the "number of
steps done". You'll find that after a certain number of rounds (most
likely just two, maybe three if you're lucky) the cost of appending
these parameters exceeds the benefit of performing additional rounds.
You can easily prove me wrong: submit a winning entry to the Calgary
Challenge. This should be trivial, since your subject line indicates
that you already have a working implementation.
DES
--
Dag-Erling Smørgrav - des@des.no
|