Takehiro Tominaga schrieb am Mon, 03 Apr 2000:
> Hi Mark,
>
> >>>>> "M" == Mark Taylor <[EMAIL PROTECTED]> writes:
>
> M> It is a little confusion to me because short blocks have fixed
> M> values for the region0/region1 partition which says that
> M> region1 ends at 576. And thus you could assume that big_values
> M> has to be at least 576 - but I just checked the mpglib source
> M> and it specifically allows big_values to be smaller than both
> M> region0 and region1.
>
> Uum,,,,, it looks you are right, Mark....
> But it is too hard to implement the count1 region quadruples compression
> into current ISO based code like lame....
>
> I make a quick hack and do it, but it is not clever and cool:
> 1 added the routine that reorder the MDCT coefficiences of short block
> 2 rewrite calc_xmin, calc_noise, count_bits, ... for short block
> 3 rewrite ShortHuffmancode routine to write out count1 region.
Is it really necessary to rewrite calc_xmin and calc_noise? What's missing
there?
> The new code still has little bug, but the snapshot will be on my web
> page(maybe tomorrow?).
>
> And bacause the new short block huffman coding algorithm is very
> efficient, I think we should change the bit allocation for short block.
A change in the bit allocation for short blocks is also necessary for
MPEG2 and MPEG2.5. At 8 kbits a short block can allocate more bits, than
possible to get (you get negative stuffing bits!).
>
> I'll check them and merge into the CVS.
> ---
> Takehiro TOMINAGA // may the source be with you !
Robert
--
e-mail: [EMAIL PROTECTED]
homepage: http://linux.unixcity.de/catwalk/index.html
--
MP3 ENCODER mailing list ( http://geek.rcc.se/mp3encoder/ )
