[Mplayer-cvslog] CVS: main/libmpcodecs vf_fil.c,NONE,1.1 Makefile,1.109,1.110 vf.c,1.92,1.93

D Richard Felker III dalias at aerifal.cx
Thu Oct 23 23:09:38 CEST 2003


On Thu, Oct 23, 2003 at 10:41:41PM +0200, Michael Niedermayer wrote:
> Hi
> 
> On Thursday 23 October 2003 22:05, D Richard Felker III wrote:
> > On Wed, Oct 22, 2003 at 11:22:29PM +0200, Michael Niedermayer CVS wrote:
> > > Update of /cvsroot/mplayer/main/libmpcodecs
> > > In directory mail:/var/tmp.root/cvs-serv14387
> > >
> > > Modified Files:
> > > 	Makefile vf.c
> > > Added Files:
> > > 	vf_fil.c
> > > Log Message:
> > > fast deinterleaver (= il filter with stride/width/height tricks)
> >
> > Huh? This code looks totally broken.
> yes, its not strictly correct, (it needs to know the stride during config() 
> ...)

Hmm, ...

Could we just wait to call config until the first put_image? Is this
valid?

> > Duplicating the il filter is not
> > possible with just stride arithmetic. 
> not exactly, but something very similar is
> 
> > It looks to me like you're
> > trying to duplicate the tfields and tinterlace filters, but not doing
> huh?!

Never mind. :)

> > it correctly... At least explain what the code is supposed to do...
> ok
> 
> input  -> (vf fil)
> 1111111  1111111XX2222222
> 2222222  3333333XX4444444
> 3333333
> 4444444
>        <- (vf fil=i)
> the XX is whatever is between width and stride/bpp
> 
> so fil splits the fields and puts them left-right instead of top-bottom

Ahh...that makes more sense! :) It's also much more useful to me,
since I like to use vf_il for watching telecined video with 2 fields
side-by-side when debugging inverse telecine code.

Rich



More information about the MPlayer-cvslog mailing list