[MPlayer-dev-eng] Re: New inverse-telecine filter

Zoltan Hidvegi mplayer at hzoli.2y.net
Thu Dec 4 10:21:14 CET 2003

> >>From the manpage I was never able to figure out what dint does, and I
> >was lazy to look at the code.  But from what you describe, this does
> >not seem to be enough, if you drop one of the interlaced frames, it
> >still leaves you with an other, and you are actually discarding a
> >field, so how can that look good?  You have to merge the last field of
> >the discarded frame with the next field, does dint do that?
>     No, it doesn't. May be I'm wrong but I've found that frames on DVDs
> (at least from filter's point of view) are progressive interlaced so I
> didn't find a way to split one frame in two fields, drop one and join
> other field with field from next frame later. Again, I may be wrong in
> that. I just didn't find another way then (a year ago). I've also tried
> detc filter but I've found it worse than dint filter because it leaved
> more interlaced frames and it looked bad. :(

What you describe is a 2-2-3-3 telecine pattern.  I've seen such a
beast, but it is very rare, over 90% of the DVDs as well as any film
source transferred to NTSC are either soft-telecined with MPEG flags,
or 2-3-2-3 telecined.  If you really have a 2-2-3-3 DVD, than you are
right, detc will not handle that, since detc tries to find a 2-3-2-3
pattern and stick with it.  For this very special case, your filter
will probably work, but for most DVDs you must do field merging, there
is no other way.  The pullup filter as well as my new filter can
recognize and handle any telecine pattern.


More information about the MPlayer-dev-eng mailing list