[FFmpeg-user] (de)-interlacing question

[Toerless Eckert]

On Mon, Nov 21, 2016 at 11:00:59PM +0100, Moritz Barsnick wrote:
> On Mon, Nov 21, 2016 at 22:32:39 +0100, Toerless Eckert wrote:
> > 2. The 50i material i get from TV nowadays seems to be 720 at 25p material
> >    that was motion interpolated to 50p, scaled down and then sent interlaced.
> >    Aka: every second field is probably just invented by motion interpolation.
> >    Which is quite annoying because IMHO this makes 50% of the bits mostly
> >    fabrication instead of actual source bits.
> 
> (Well, if they don't have intermediate frames originally, but they do
> have bandwidth to spare, you can't blame them.)

I can and i do blame them. I wish there was more of an "originalist" thinking
where you avoid all unnecessary messing with the media. I understand why
they do it though. Its actually quite self-serving. It maximizes the visual
experience on the oldest of receivers while simultaneously reducing its
actual quality and making it most difficult to deal with it in modern
receivers and ultimately allowing them to have the best arguments to switch
off SD as soon as possible.


> Speaking for German DVB-S2 with 720 at 50p: If you look at that material
> frame by frame, you will notice that it's not actually interpolated,
> but instead frames are just repeated (when the input material was
> recorded at 25p). After inspection of a recording, you can just insert
> "-vf fps=fps=25". It won't make your target video *much* smaller
> though, as repeated frames probably compress quite nicely (with H.264),
> but you will notice the difference in size. And you won't notice a
> difference in motion. ;)

Exactly. That was my analysis as well. And then i checked and compared with
the SD PAL (DVB-S) simulcast of the same materials. And figured how they
do this crappy motion interpolation to get to 50i.

Alas, i still need to record mostly SD. But yes, this crappy interpolations
has me evaluate if it would be easier to fix my HD playout toolchain  issues
than solving the deinterlacing issue in an acceptable fashion.

Toerless

> Moritz
> _______________________________________________
> ffmpeg-user mailing list
> ffmpeg-user at ffmpeg.org
> http://ffmpeg.org/mailman/listinfo/ffmpeg-user
> 
> To unsubscribe, visit link above, or email
> ffmpeg-user-request at ffmpeg.org with subject "unsubscribe".

-- 
---
tte at cs.fau.de