[mythtv] HDTV patches

Isaac Richards ijr at po.cwru.edu
Fri Apr 30 10:45:01 EDT 2004


On Friday 30 April 2004 08:58 am, Daniel Thor Kristjansson wrote:
> That change is required for frame skipping. You need to keep track of
> both the frames actually shown on the screen vs the frames in the movie.
> If you don't skipping backwards and forward doesn't work. The NVP
> keeps a ticker of the frames shown, so if it had only shown 90% of the
> frames after 20 minutes it's time in recorded frames would be 2 minutes
> off.

If that were so, why are:

void SetRecorderFramesPlayed(long long played) { recorder_frames_played = 
played; }
void SetFramesPlayed(long long played) { recorder_frames_played = played; }

identical?  And:

-    long long framesPlayed;
+    long long recorder_frames_played;

And why is it called 'Recorder' when the recorder has absolutely nothing to do 
with it?

> I will annotate all the changes in NVP so that you can see the logic in
> it.
>
> ]Actually, check that.  Just finished testing the xvmc parts, and, it
> doesn't ]seem to do so well -- picture quality doesn't look all that great
> (much more ]pixelated than previously), and it dies pretty horribly on one
> of my test ]clips (video size changes 4 times in a row, 10 or so seconds on
> each ]resolution).  So, not ready for CVS just yet, I'd say.  =)
> Hmm, can you point me to that video clip?
>
> The pixelation is because it tries to figure out if the frame might be
> interlaced, and if thinks it is (such as for 640x480i clips) it only
> displays half the rows. It's a poor man's deinterlacing. I've only
> noticed the pixelation from this in the OSD, but I can disable this
> feature for 640x480i so the behavior doesn't change for PVR-x50 users.
> At 1080i I think the loss in vertical resolution is compensated by the lack
> of interlacing artifacts, at 720p it doesn't do deinterlacing. So I
> think 480i is the problem resolution. I have few 480i clips myself.

I tested 1080i, 480i, 480p, and 720p.  All looked much worse with your patch 
than before.

Isaac


More information about the mythtv-dev mailing list