[mythtv-users] 1080i to 1080p Deinterlacing on Backend was Raspberry Pi now ships with 512MB RAM

[Scott Knight]

on 10/17/2012 8:05 AM Andrew Theurer carved the following into a picnic 
table:
> On Tue, Oct 16, 2012 at 10:34 PM, Scott Knight <scott at scottknight.com> wrote:
> > on 10/16/2012 11:00 PM Michael T. Dean carved the following into a picnic
> > table:
> >>
> >> And, yes, a system that's transcoding everything to H.264 (versus
> >> storing it in its native recording format) does make for an extremely
> >> power hungry system.
> >>
> > You have just touched on something I have looked at a few times and never
> > successfully figured out.  I already have a powerful backend (Xeon E3) with
> > power to spare because it does other things like hosting VMs for building
> > MythTV, cloud testing, etc.  It can commflag at 1200+ fps on 1080i content.
> > What I have noticed is that 1080i content looks pretty crappy no matter
> > which deinterlacer I use on the frontends (mixture of ION, ION2, 9400, GT430
> > all running VDPAU).  720p content looks better and Blu-Ray looks stunning
> > without even making the frontends break a sweat.
>
> Scott, are you sure the more advanced de-interlacers are being used,
> especially on the GT430 card?  I have used the 2x temporal, and I
> thought it looked quite good.  This is a on a very large screen
> (projector), which tends to show imperfections quite easily.
Trying the different settings used to be quite tedious, but I discovered 
with 0.25 that there is a handy menu that lets me switch them on the 
fly.  Makes for fun back to back to back evaluations. The irony here is 
that in our main family room with the 'big' screen, the screen isn't so 
big (46") and we sit quite far away (~17 feet).  That is the FE with the 
GT430 that feeds a Denon 4311ci, which by all accounts has a pretty 
sweet conversion chip that I have never tried.  The only reason for 
switching to the GT430 was to get the HD audio.  Honestly, I think the 
deinterlace settings might be default and since we sit so far away, have 
never had a reason to muck with them.

We do view some of the other screens much more closely, though, and this 
is where I see the interlace artifacts.  Some of the interlace modes 
look decent, but none look as good as a progressive stream. Maybe I am 
just doing it wrong.
> > Is there a way to take my 1080i recordings and just deinterlace them on the
> > backend?  I don't care about transcoding out the commercials, don't care
> > what format they end up in, don't care about disk space, as long as I can
> > use whatever deinterlace algorithm looks best to my eyes when it's played.
> > Maybe I don't understand the limitations of 1080i deinterlacing, but it
> > seems that with a good enough transcode, the end result should be somewhere
> > between 720p and Blu-Ray.
>
> That got me thinking, since a lot of prime time shows are actually 24p
> that are telecin'd to 60i, I wonder if you could convert to the
> original 24p source.  My preference, however, is to never transcode,
> as I find it to always degrade the picture quality.
How would I go about interrogating the file to see that?  My files 
pretty much look like this:
Input #0, mpegts, from '/mythtv/recordings/6858_20121014195900.mpg':
   Duration: 01:00:51.39, start: 55639.415922, bitrate: 12437 kb/s
   Program 1
     Stream #0:0[0x18b1]: Video: mpeg2video (Main) ([2][0][0][0] / 
0x0002), yuv420p, 1920x1080 [SAR 1:1 DAR 16:9], 80000 kb/s, 29.97 fps, 
29.97 tbr, 90k tbn, 59.94 tbc
     Stream #0:1[0x18b2](eng): Audio: ac3 ([129][0][0][0] / 0x0081), 
48000 Hz, 5.1(side), s16, 384 kb/s
     Stream #0:2[0x18b4]: Unknown: none ([5][0][0][0] / 0x0005)

or this:
Input #0, mpegts, from '/mythtv/recordings/5583_20121014210000.mpg':
   Duration: 01:00:59.11, start: 89768.463944, bitrate: 13972 kb/s
   Program 1
     Stream #0:0[0x1c71]: Video: mpeg2video (Main) ([2][0][0][0] / 
0x0002), yuv420p, 1280x720 [SAR 1:1 DAR 16:9], 18000 kb/s, 59.96 fps, 
59.94 tbr, 90k tbn, 119.88 tbc
     Stream #0:1[0x1c72](eng): Audio: ac3 (AC-3 / 0x332D4341), 48000 Hz, 
5.1(side), s16, 384 kb/s
     Stream #0:2[0x1c73](spa): Audio: ac3 (AC-3 / 0x332D4341), 48000 Hz, 
stereo, s16, 192 kb/s
     Stream #0:3[0x1c74]: Unknown: none ([134][0][0][0] / 0x0086)

Maybe I am barking up the wrong tree here, but it just seems to make 
sense that if the stream has to be deinterlaced at some point before the 
photons hit my eyeballs, I could leverage the heavy duty backend to do 
the best job possible.  I am going to experiment a little with Pau 
Gardiner's command.

Thanks, Scott