[MPlayer-users] [BUGREPORT] pullup/filmdint bugs when height/width %16 != 0

Rich Felker dalias at aerifal.cx
Mon Aug 1 16:58:29 CEST 2005

On Mon, Aug 01, 2005 at 02:41:07AM -0700, RC wrote:
> On Sat, 30 Jul 2005 10:03:40 -0400
> Rich Felker <dalias at aerifal.cx> wrote:
> > The issue is not multiples of 16, but multiples of 4, and height only.
> I see.
> > Interlaced 4:2:0 video must always have a height that's a multiple of
> > 4; otherwise it's nonsense.
> Of course I understand that (you don't pay much attention to who sent
> messages, do you?).  
> All of the videos I had problems with were progressive...  pullup and
> filmdint (with io=1:1) shouldn't munge progressive material, right?

Yes but they still have to consider it as interlaced for processing.
And this requires having two complete equal sized fields. I mean I
could make it add a black line to one field if the size is bad, but
this seems silly..

> Also, even though it doesn't strictly make sense, I would think you'd
> want ivtc and deinterlacer filters to work reasonably well, even on
> material that was encoded by someone who didn't know what they were
> doing.

You can crop if necessary.. These filters aren't meant to be used on
random encoded files you find on the net, but canonical streams from
tv broadcast or dvd..

> > I don't know why MPlayer crashes; 
> Well, that's why I filed the bugreport...
> > P.S. About that scaling... I hope you're doing interlaced scaling. :)
> I'm not actually using the scaling method for encoding, just for
> live playback with MPlayer.  I've only seen this bug with progressive
> material so far, anyhow.

Yeah, interlaced material will never have non-mod4 height, at least as
long as it's from a legitimate source (tv broadcast, etc.). If some
moron captured it, all bets are off, but in that case they probably
already ruined the interlaced video with scaling, bttv yv12 mode,
deinterlacing, etc. anyway..


More information about the MPlayer-users mailing list