[MPlayer-users] new version of detc filter! (inverse telecine)

[D Richard Felker III]

On Sun, Mar 16, 2003 at 01:43:46AM +0100, Arpi wrote:
> [Automatic answer: RTFM (read DOCS, FAQ), also read DOCS/bugreports.html]
> Hi,
> 
> > > > DRFI>Overall, this code seems to be working much better on my test
> > > > DRFI>material (still SE Lain). At this point I'm mostly happy with it,
> > > > 
> > > > Hmm, when I play SEL encoded to svcd I see almost no interlace frames,
> > > > but mplayer (in demuxer_mpg part) keeps on switching framerate
> > > > ("detected telecined sequence...", after a while "detected progressive
> > > > sequence...", and so many times), like it was without detc in recent
> > > > mplayer. Isn't this behavior interfering with detc filter, I wonder?
> > > 
> > > it does, imho.
> > > I think that detc filte should check the telecine flag (set by mpeg demuxer)
> > > and disable itself while it's playing telecined frames.
> > 
> > How???
> 
> video.c:
> -static int telecine=0;
> +int telecine=0;
> 
> :)
> 
> and don't drop frames when it's non-zero
> currently your filter drops every 5th frame even if all the frames are
> non-interlaced.

Only if dropping is enabled. The idea is that you want to output
constant-framerate video, and the filter assumes the input framerate
is constant. Progressive sequences are assumed to be 30fps progressive
rather than 24fps, i.e. it's made for dealing with files that are all
encoded at 30fps but which might be a mix of real 30fps sequences and
24fps sequences telecined up to 30. However, dealing with badly
encoded movies which mix hard and soft telecine should be possible
with the telecine var from video.c.

While this is possible, IMHO a much better approach would be to put
timestamps in mpimage_t, and let filters change them when passing the
image on to the next layer. This will eventually facilitate proper
framerate-changing filters without all the stupid hacks I've done so
far, and they'll work better with mencoder!

Rich