[MPlayer-users] -vf ilpack

[D Richard Felker III]

On Mon, Mar 08, 2004 at 04:36:30AM +0200, Ville Saari wrote:
> On Sun, Mar 07, 2004 at 01:52:07PM -0500, D Richard Felker III wrote:
> 
> > In fact in some cases it might be optimal to encode interlaced frames
> > without any of the interlaced features (e.g. if there's almost no motion)
> 
> True, but for such frames it wouldn't hurt if progressive chroma subsampling
> is used too.

The way chroma sampling really works, you're right. However, if it
worked the stupid way I used to think it did, bad things would
happen...

> > > Several PAL DVD are encoded so that progressive film content was
> > > converted to interlaced video, which was then encoded to DVD with
> > > opposite field dominance so that every frame is like the interlaced
> > > frames of telecined NTSC.
> > 
> > Uhg. Do you have any idea what this does to the quality/bitrate?? :(
> 
> If the mpeg encoding was done using the interlaced features, then the
> quality loss doesn't seem to be too bad, but if it was encoded

Even with interlaced features, the quality loss is huge. You lose a
huge part of the benefit of motion estimation.

> I have also witnessed at least one case where 24 fps film content was
> converted to PAL using 3:2:2:2:2:2:2:2:2:2:2:2-pulldown! The result
> is visible jerks with constant 2 Hz frequency. Looks particularly
> terrible when the camera is panned.

Well normally this is done for movies where the music is the key
feature, and increasing the pitch by 4% would butcher it. I don't
really know any good way to handle this situation... :(

> > Applying the linear blend filter afterwards should clean it up enough
> > to be watchable without hurting the video quality too much. It will
> > halve the ghosting in each field and spread both ghosts out over both
> > fields so you don't see combing.
> 
> Yes, but vertical resolution is always compromised if the video is
> deinterlaced. At least until someone implements a motion compensating
> deinterlecer to mplayer :-)

Linear blend doesn't kill vertical resolution. It does give slight
blurring, but it's far from a dimension-halving filter like li,ci,fd.

> I personally prefer pp=md over the other deinterlacers. Those deinterlacers

Yes, md is good. But it will necessarily destroy all local extrema in
the image, which looks bad. A slightly better algorithm is needed for
choosing when to replace a pixel and what to replace it with.

> In this particular case pp=md banishes one of the ghosts completely and
> removes the combing from the other.

This still sounds bad... Care to write a nice filter to detect the
ghosts from the previous frame and remove them? :)

Rich