R: [M4IF Technotes] Questions on H26L

[Christoph Lampert]

On Thu, 15 May 2003, Lucia wrote:
> Hallo Gary,
> 
> I'd like to thank you for your answer, it's been quite clarifying to me.
> I guess the overall bit-rate constraint (Rc) is calculated in someway like
> [(channel_bit_rate/frame_rate)-bits for already encoded MBs in the frame)],
> is it?
> 
> Still, the relation between the multiplier and QP is showed to be quadratic
> in most of the papers i've read explaining the tie itself, but in the H26L
> codec a formula
> with an exponential and a squared term is used... is it an adjustment
> deriving from empirical
> experiments with different possible ties?

It's the same "value" for lambda (quadratic in quantization step size),
but in H.26L this step size doesn't depend linear on QP anymore,
but exponentally in this way. 
Christoph 
-- 
Dr. Christoph H. Lampert chl   math,uni-bonn,de  | Diese Signature ist maschi-
Beringstr. 6, Zi. 15, 53115 Bonn, Germany      | nell erstellt und auch ohne
Tel. (0228) 73-4708  Fax. +49 228 73-7916      | Unterschrift wirksam.
http://www.math.uni-bonn.de/people/chl/        |                    AZ 27B/6