No subject

Wed Jun 24 15:02:31 UTC 2009

myself that the shows that MythTV (and mplayer) have problems with are
encoded with soft telecine.  Eg, shows from stations here which play
back fine (PBS, CW) don't show repeat flags, but shows from stations
here that give MythTV problems (CBS, NBC) do show repeat flags.

So it seems that the simpler ecoders will do hard telecine, and the
more sophisticated encoders will do soft telecine.  My supposition is
that most software players (mplayer, MythTV) simply cannot correctly
detect soft telecine in HD broadcasts, and repeatedly switch beteen
interlaced and progressive modes every few frames.

> problematic flagging does not come from the source material, as its
> existence varies by show and by episode, apparently determined by
> contemporary bandwidth-shaping needs on the broadcaster's part; I have
> seen many cases in which HDNet Movies will show a given movie without
> issues, unlike (say) Showtime HD.

Again, I think it is an issue of the sophistication of the encoder.
HDNet is probably using an older encoder, or one with different
defaults than Showtime-HD.

> > From what I understand, its very uncommon for software handle this
> > correctly.  I'm just hoping VDPAU will play these files back
> > correctly..
> The issue is unrelated to VDPAU or to graphical horsepower, or the
> lack thereof. VDPAU offloads to the GPU the graphical decoding,
> rendering, and deinterlacing duties that, for MythTV, the CPU has
> traditionally handled. That's it. Yes, VDPAU's deinterlacers are
> extremely good (both in and of themselves, and in the sense that the
> offloading often permits the use of most-sophisticated deinterlacers
> than would be possible otherwise). VDPAU does not, as far as I can
> tell, make a meaningful difference in terms of this particular issue,
> however, as it arises solely from MythTV frantically switching between
> deinterlacing and not deinterlacing multiple times a second as it is
> told to. The aforementioned patches simply tell mythfrontend to ignore
> progressive-frames flags during playback once it sees an interlaced
> flag; in other words, ignorance is bliss from the MythTV
> perspective.[1]

As long as VDPAU offloads the decoding, and not just the rendering,
then I think I'm sold.  As you said, its easy enough to force
MythTV to not flip/flop.  I was worried that some of the decoding
was still managed in software.

> You asked about video quality comparable to that of a Sigma-based
> Network Media Tank. I have never owned one so cannot say firsthand;
> someone like Brian Wood would be able to compare. I am more than
> satisfied with the performance of VDPAU's Temporal 2x deinterlacer,
> but Jean-Yves Avenard and others have spoken of the Advanced 2x
> deinterlacer (which my 8400GS video card can't handle) as being even
> better.

I'm still reading up on which deinterlacers and things that
the Ion supports. It looks like it can't quite do Advanced 2x
for 1080i.

> [1] Now that I think about it, I suspect that NMTs and other settop
> boxes that don't have issues with the broadcasts in question simply do
> the same thing as the patches do, and ignore the flags when they don't
> make sense.

I imagine they actually do correct decoding of 1080i..  At least my
Sigma based SageTV HD100 looks *far* better than anything I've ever
seen MythTV (or mplayer) do when playing back 1080i from CBS and NBC.
Its just that the interface is painfully slow to respond, I'm unhappy
in general with SageTV, and so I'm looking for alternatives..


More information about the mythtv-users mailing list