[Mplayer-cvslog] CVS: main/libmpcodecs vf_scale.c,1.16,1.17

Michael Niedermayer michaelni at gmx.at
Sat Jul 6 01:12:34 CEST 2002


Hi

On Friday 05 July 2002 22:30, Felix Buenemann wrote:
> On Thursday 04 July 2002 15:08, Michael Niedermayer wrote:
> > Modified Files:
> >         vf_scale.c
> > Log Message:
> > more scaling algos
>
> hmm, nice.
>
> I noticed, that the Sinc algo seems buggy, I got brighter vertical lines,
> when scalong a 348x172 DivX (janwayne.avi) to 960x474. (the avi should be
> somewhere on mphq)
hmm, the sinc filter is pretty long & slow so i think (could be wrong though) 
that this is caused by rounding or something, anyway sinc looks bad (ringing 
around sharp edges) so it isnt very important ...

>
> Btw. could you write some description of the different algos and their
> pro/cons? (eg. up/downscale qulity and speed).
:(

>
> Btw2. what's the experimental scaler again= it looks a bit like a mix
> between bilinear and point scaler.
... experimental ... and that one is a failed experiment ;)
the filter is pretty much the middle part of the cubic filter scaled by pow(), 
the idea was to create a filter which is shorter (=faster) than the cubic one 
and looks better than bilinear ... this obviously was a stupid idea as a 
filter with only 2 taps like bilinear must be the bilinear filter or it will 
trash even a constant gradient -> thats why it looks like the point scaler

Michael



More information about the MPlayer-cvslog mailing list