> From: Mathew Hendry <[EMAIL PROTECTED]>
> Date: Thu, 20 Jan 2000 14:30:52 -0000
> Maybe I'm on completely the wrong track here, but could they be recomputing
> masking thresholds after each quantization?
>
I would guess it is slower because it does a more thorough
search for the best scalefactors. Checking scalefactors
(one iteration of outerloop) is very slow. And this
would be consistent with the description of the quality options
in FhG.
Take a look at typical scalefactors from lame or xing and compare them
to FhG: FhG really exploits the full range of possible scalefactors,
and lame clearly does not.
I dont know of any models for masking which take into the account
the amount of noise that will be introduced into the masker. But
FhG surely knowns many things we dont!
Mark
--
MP3 ENCODER mailing list ( http://geek.rcc.se/mp3encoder/ )
