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

[D Richard Felker III]

On Thu, Dec 04, 2003 at 05:20:41PM -0600, Billy Biggs wrote:
> Billy Biggs (vektor at dumbterm.net):
> 
> >   I also won't believe you if you say that doing the progressive
> > downscale on progressive content will be _higher_ quality.  Keeping
> > chroma from only one field seems like a bad, bad idea, given how NTSC
> > encoding works.  I bet that if you did a proper upsample converting each
> > field to 4:2:2, then merged them, then downscaled using a good filter
> > back to 4:2:0, you would get a higher quality image.
> 
>   I think I can make this argument even stronger.  Say I have a 4:2:2
> frame and it IS progressive.  Is it correct for me to DROP every second
> scanline of chroma to get a 4:2:0 frame?  No, it is not.  I should
> downsample using a higher order filter, even better, one that's actually
> to spec for MPEG2's 4:2:0 format.  Otherwise, I get aliasing.
> 
>   So clearly, just looking at chroma from _one_ field and calling that
> the chroma plane of a 4:2:0 image is never correct.

I agree totally. The solution is simple. V4L drivers should never
report that they support YV12 unless the _hardware_ is capable of
filtering in such a way that you get correct results for both the
interlaced and progressive case. (Actually it could be done in
software too, but video filters do NOT belong in the kernel!)

Perhaps they could report that they support YV12 only for single-field
capture modes (since these are used only for low-quality,
quick-and-dirty tv watching anyway), but I imagine the V4L api sucks
too much for cards to have different capabilities for different
capture sizes...

Rich