On Sun, 15 Aug 2004, Florian Schaefer wrote:
> whole movie, only the first seconds for testing purposes. But the
> beginning seems to work with mplex:
> INFO: [mplex] Peak bit-rate : 8694400 bits/sec
Ah, ok. That's fine. One some material the peaks (which can be very
brief) will be considerably higher than the average - if the peak is
higher than the maximum for a DVD then hardware players will have
a problem playing the disc.
> > You can determine if this is the problem by adding the option
> > "--no-dualprime-mpeg2" to the mpeg2enc command. The CVS version has
>
> Hmpf, my 1.6.2 doesn't know about this option. Even "strings" cannot find
> the string in the binary.
Oh ok, then that was added after the release of 1.6.2
> Therefore I just compiled the CVS version. However, the results do not
That's good - there have been numerous small fixes, enhancements,
and overall cleanup done done. One big change was a new and improved
YUV4MEG2 API.
> look different, even if I use the above option with the CVS mpeg2enc.
;(
> > The other possibility is that you're driving the DCT/iDCT into overflow
> > with -q 3. Does the artifacting go away if you use "-q 4"?
>
> That seems to have been the case. If I use "-q 4", the artifacts seem to
> go away. :-) Thanks for this tip.
Super! You're welcome. On the IA32 platform it seems that -q 3 is
often too "aggressive" and can cause the MMX/SSE routines to
experience an occasional overflow. The PPC Altivec routines do not
have this problem from what I have seen (I haven't seen the problem
on my G4 or G5 OS/X systems)
> [...]
> > mpeg2dec -o pgmpipe movie.m2v | pgmtoy4m -i t -r 30000:1001 -a 10:11 | \
> > mpeg2enc -q ... -o movie_new.m2v
> >
> > You may need to change the '-i t' to be '-i b' if the video is
> > from DV (which is always bottom field first0 and the rate (-r) to be
> > 25:1 for PAL, and so on.
>
> Great. So for my PAL DV material I use "pgmtoy4m -i b -r 25:1 -a 3:4".
Aiieee ;)
The aspect is the SAR (Sample Aspect Ratio) not the DAR (Display Aspect
Ratio). For PAL 4:3 material the SAR is 59:54 (NTSC is 10:11). Thus
the aspect arg should be "-a 59:54" for PAL. Widescreen PAL would have
use "-a 118:81". See http://www.mir.com/DMG/aspect.html
> Having hopefully settled this "error" I can now turn my attention to my
> second problem: Interlacing (I've come to loathe it).
I'll agree with that! Like a lot of things in the video world we
are still paying for the technology limitations from ~60 years ago.
Interlacing, 24fps, and others are all derived from working around
the limits of when TV was being invented.
> I've tried bottom- and top-first encoding but the DVD player just keeps
> on showing very jerky motions. Either it thinks that it is a progressive
> stream or it uses a wrong field order, but sadly the player doesn't
That's why I do NOT use mplayer's yuv4mpeg option - the YUV4MPEG2
header is often not correct enough to pass thru into a mjpegtools
pipeline. If you fix the header the decoded data is right, but
often the frame rate and interlacing parameters are wrong (the sample
aspect may also be incorrect). So, I created the small utility
'pgmtoy4m' to handle mpeg2dec's "pgmpipe" format.
> PS: Is the CVS mpeg2enc version quicker than the 1.6.2 release? I'm
> noticing encoding speedups around 50%. That's great. :-)
Yes, it is. I think that after the release the default was changed
to not use B frames which take longer to compute than the P frames.
If you want 1 or 2 B frames between P frames you can use "-R 1" or
"-R 2". So far all the encodings I have done show a lower average
bitrate if B frames are NOT used.
Cheers,
Steven Schultz
-------------------------------------------------------
SF.Net email is sponsored by Shop4tech.com-Lowest price on Blank Media
100pk Sonic DVD-R 4x for only $29 -100pk Sonic DVD+R for only $33
Save 50% off Retail on Ink & Toner - Free Shipping and Free Gift.
http://www.shop4tech.com/z/Inkjet_Cartridges/9_108_r285
_______________________________________________
Mjpeg-users mailing list
[EMAIL PROTECTED]
https://lists.sourceforge.net/lists/listinfo/mjpeg-users
