[Ffmpeg-cvslog] r8182 - trunk/doc/TODO

Michael Niedermayer michaelni
Fri Mar 2 14:25:58 CET 2007


Hi

On Fri, Mar 02, 2007 at 02:16:31PM +0100, Baptiste Coudurier wrote:
> Hi
> 
> Michael Niedermayer wrote:
> > Hi
> > 
> > On Fri, Mar 02, 2007 at 10:54:19AM +0100, Baptiste Coudurier wrote:
> >> Hi
> >>
> >> Nico Sabbi wrote:
> >>> bcoudurier wrote:
> >>>
> >>>> - mpeg2 non linear quantizer
> >>>>  
> >>>>
> >>> is this implemented? how to use it? what are the benefits?
> >>>
> >> Nope, It's on my TODO, after digging a bit, it seems it would be needed
> >> to use a completely different quantization code, since other codecs
> >> using mpegvideo.c are using linear one, maybe Im wrong though, some mpeg
> >> code is still black magic for me.
> >>
> >> static const uint8_t non_linear_qscale[32] = {
> >>     0, 1, 2, 3, 4, 5, 6, 7,
> >>     8,10,12,14,16,18,20,22,
> >>     24,28,32,36,40,44,48,52,
> >>     56,64,72,80,88,96,104,112,
> >> };
> >>
> >> Benefits is capability to use 1,3,5,7... quantization coefficients, and
> >> then going up to 112, which in practice improve very high bitrates
> >> quality and also should compress more at (very)low bitrates.
> > 
> > video above mpeg2 qscale 62 looks too bad to be useable, and below 
> > mpeg2 qscale 2 you will not see a difference, the only usefull thing
> > in this is maybe 3 and 5
> 
> yes this is not useable, sometimes useful for dirty screening at very
> low bitrate for equipment only supporting mpeg2...
> 
> 1,3,5,7 quants offers actually better quality at high bitrates,
> you gain granularity. I compared it much using mainconcept encoder,
> but yes this is debatable.
> 
> > so IMO the whole thing is a waste of time
> 
> Well, non linear q is mandatory for D10, since D10 is useful for me, non
> linear q is.

if(non_linear_qp){
    assert(qscale>=1 && qscale <=12);
    put_bits(inv_non_linear_qscale[qscale];
}else{
    put_bits(qscale);
}

and a 
if(non_linear_qp && qmin >12){
    av_log("not supported blah blah");
    return -1;
}
somewhere

and you have non linear qp support


[...]
-- 
Michael     GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB

The educated differ from the uneducated as much as the living from the
dead. -- Aristotle 
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: not available
URL: <http://lists.mplayerhq.hu/pipermail/ffmpeg-cvslog/attachments/20070302/febfb2e4/attachment.pgp>



More information about the ffmpeg-cvslog mailing list