[vorbis] RC2 vs. beta4 (test)

[Monty]

On Fri, Aug 24, 2001 at 09:00:49AM +0200, Attila Padar wrote:
> Hi again!
> It seems the bug is in the noise masking of RC2,
> because if I turn off the noise masking in the 
> MODE_44_C.H , I don't get the same bug after 
> the 10. encoding either (but I get a double-size file)...
> The beta4 also uses noise masking on 256 kbit/s,
> but - it seems in the test file - it encodes the 
> sound correctly.
> I wonder, how do the developers think it:
> is this really a bug or it isn't?
> Or maybe the noise filtering is too much,
> and you should use less noise with more
> tone and peek filtering.

The noise masking algo in rc2 is different than the one in b4.  Each
had its liabilities:

b4 used a median-based floor chooser and tended to produce very
different results in long and short blocks for the same audio.

rc2 uses a least-^4th-power line fitter, which has very good agreement
between block sizes...  but is turning out to have other quirks.

To be fair, neither is the best answer as both are very sensitive to
generational reencoding and transcoding.  "I'm still working on it".
I have more faith in the rc2 system, but I need to spend more
tuning/experimentation on it.

In both b4 and rc2, practically any artifact you hear is going to be
due to noise masking.  It's a necessary evil.

Monty

--- >8 ----
List archives:  http://www.xiph.org/archives/
Ogg project homepage: http://www.xiph.org/ogg/
To unsubscribe from this list, send a message to 'vorbis-request at xiph.org'
containing only the word 'unsubscribe' in the body.  No subject is needed.
Unsubscribe messages sent to the list will be ignored/filtered.