[mythtv] packed image format and new deinterlacers

[Markus Schulz]

Am Samstag, 16. September 2006 01:13 schrieb Doug Larrick:
> Markus Schulz wrote:
> > Perhaps this would clear with the following:
> >
> > Currently i got: (25FPS for PAL)
> > Plugin Call: Frame with Field 1 + 2
> > Plugin Call: Frame with Field 3 + 4
> >
> > Nice would: (50FPS)
> > Plugin Call: Frame with Field 1 + 2 (top first)
> > Plugin Call: Frame with Field 2 + 3 (bottom first)
> > Plugin Call: Frame with Field 3 + 4 (top first) ...
> >
> > It would be nice to hear some comments on this, i would like to put
> > some of the DScaler deinterlacers into mythtv. But at least the
> > frame history is a must have for this.
> >
> > Any Suggestions?
>
> The main problem you'll run into with nicer deinterlacing is that a
> filter has no way of doubling the frame rate, except for the hack I
> used when I wrote the current bob filter (where collusion between the
> filter and the video-out method shows alternate halves (top & bottom)
> at double the frame rate).  There's no way to get full-sized, 
> double-frame-rate, frames out.  I'm sure it would be possible to fix
> this, but will require changes to more than just the filters.

hmm ok, sounds like a lot of work for me ;)

> If you do manage to fix this, what I might suggest to get the data
> more advanced filters need is to eat the first three frames, storing
> them in whatever format is convenient, and output black replacement
> frames. Then when you have enough frames you can start outputting

but with delaying the first frames i should go into AV-Sync troubles 
which should be noticable by the user.
if i solved the above i would start with a first "copy-needed-frames" 
version. 

> real data.  If you're going to go to this trouble, you might adapt
> the filter interface to one used by some standard tool, such as
> dscaler or tvtime, that has a variety of filters available.

this sounds better, but at the cost of really much work. 

Currently i'm searching for the source where the fields will weaved 
together into the AVFrame (and later VideoFrame). Still not found yet.

 
Markus Schulz