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

[Rich Felker]

On Sat, Jul 23, 2005 at 07:41:06PM -0700, RC wrote:
> On Sat, 23 Jul 2005 10:48:57 -0400
> Rich Felker <dalias at aerifal.cx> wrote:
> 
> > No, I have never said this!!!!!!!!!!!!!!!
> > I said that all of my inverse telecine filters will do no damage to 30
> > FPS PROGRESSIVE CONTENT! They will all butcher interlaced content in
> > random ways. The only inverse telecine filter 'safe' to use on
> > interlaced content is filmdint, which was not written by me.
> 
> Here's exactly what you said:
> 
> > detc will 'probably' pass all fields thru unmodified when it doesn't
> > see any telecine pattern, but if there are a few fields with very low
> > motion, it could mistakenly think it's the result of telecine and try
> > to process the pattern, resulting in highly broken output. the only
> > inverse telecine filter 'safe' to be used on interlaced video content
> > is filmdint, and even that is a very bad idea.
> 
> Do you wish to change or retract that statement?  

What part of "RESULTING IN HIGHLY BROKEN OUTPUT" do you not
understand? "Probably" passing fields through unmodified is not enough
when even a probability of 99% will give you over 500 errors per
30-minutes of content.

> I certainly haven't noticed negative effects from detc despite doing a
> lot of TV capture using it, but if it's only modifying fields during
> low-motion, perhaps it's just not very noticable.

The thing about detc is that it latches onto a pattern without any
access to future context. So if it sees a false pattern during low
motion, it may butcher a few high motion fields right afterwards
before shutting off. If it works ok in practice for you, that's good,
but I certainly wouldn't trust it...

Rich