[MPlayer-users] Re: [-] Re: [-] Re: [-] Crop before deinterlacing OK?

[Rich Felker]

On Sat, Jul 23, 2005 at 09:56:37AM +0200, Matthias Wieser wrote:
> > Your only 
> > hope of good results with deinterlacing is to use a 60fps
> > deinterlacer, i.e. tfields.
> 
> This will be included into an additional profile for Kalva.

:)

> > > > That's because when you ivtc, you most of times change
> > > > framerate back to mostly the original framerate of the movie. Now,
> > > > if you do not set the correct output framerate with "-ofps", you'll
> > > > either end-up with duplicate frames
> > >
> > > How does it look like? I could imagine that those CNN live tickers
> > > would show some unsteady motion if frames were duplicates. But does
> > > it matter for movies?
> >
> > Yes, especially during pans.
> 
> When will Mplayer be able to smooth the GMC vektor during playback? ;-)

No, this will only cause image corruption. Even minute changes in
motion compensation will corrupt the image significantly over
200-frame key intervals, and since video is undersampled temporally
there's really no way of fixing choppy motion with filtering even if
you can avoid this issue. You're suggesting a very high-tech solution
to a PEBKAC error.. :)

> > > But let's assume someone is recording a movie. Would detc be an good
> > > option? The manpage says that ivtc does not like noisy input.
> >
> > It really depends on how bad your TV signal is. Cable (with a good
> > connection) or satellite should be fine. Terrestrial analog broadcast
> > may be good or bad depending on the user's antenna system and the
> > proximity to the station.
> 
> Then I will choose detc as default.

OK. detc will likely butcher non-telecined content though.

> > > What would be an good option for the daily soap opera? ivtc together
> > > with -ofps 30000/1001, or better detc (any advice for the
> > > sub-options?) with -ofps 30000/1001?
> >
> > No, soaps are not telecined as far as I know;
> 
> That's what I thought, too. But see my mail from 07.07.:
> ---
> [I wrote:]
> > I thought you only need to inverse teleine when capturing NTSC movies. 
> > Normal NTSC TV broadcasts can be deinterlaced with pp=[md|lb...], right?
> [You wrote:]
> It depends on the content. You have to painstakingly check every
> single thing you encode.

:)
Yes, some high-budget TV series these days are shot on film and then
telecined. Low-budget stuff is usually shot with video cameras (60i).

> Do you think it would work for *most* (not all) live recordings to provide 
> two profiles: one with detc and -ofts=~24 and another one with kerndeint?

You'd better test and see what you think. To me, it will look
horrible, but you might find it acceptable. detc is unreliable enough
IMO that I would at least follow it with pp=lb. lb doesn't do too much
damage to progressive frames (unlike proper deinterlacers), and if
detc accidentally makes some combed frames pp=lb will hide them with
some blurring.

Rich