[MEncoder-users] Highest quality DVD -> MPEG2 rip

[percy tiglao]

I've been looking up dvd ripping and so forth. I have a couple DVDs of
a few anime that I'd like to rip onto my computer so that I can make
an AMV (anime music video). Pretty much, it'd be a fan made video
using footage from the dvds. However... because all of the videos are
going to be edited (scaling effects and so forth), I would like to
have the rip to be as high quality as possible... while having a
consistent format. Unfortunately, the various DVDs of the same anime I
have seem to be of different formats. (some are hard-telecined, others
are soft-telecined, etc. etc.)

If I could, I would just edit straight out of the VOB file. However,
I'd like to apply a pullup filter to bring all of the dvd data to a
consistent ~24 progressive fps. The -vf pullup,softskip filter is
working perfectly, but now -ovc copy won't work because I'm using a
filter. So the issue comes to re-encoding the video. (audio is not
really important).

Size is almost not an issue. I'm willing to have ~1 GB per 30 minutes
of video if it is necessary (and have reserved the hard-drive space to
do so)... but I don't have the hard-drive space for 13 hours of DVD
rips under a lossless codec. >_<.

So here are the issues I'm concerned with.

1. Does re-encoding in MPEG2 help the quality of the video? I'm
worried that converting the MPEG2 data on the DVD into XVid or
something will compound compression artifacts. Does re-encoding the
DVD data as MPEG2 prevent this (or at least hamper it)?

2. At least to my PSNR experiments... vqscale=1 seems to be working
better than vqscale=2. However, the man page seems to not recommend
vqscale=1. I searched the documentation, and there are references to
"DOCS/tech/libavc-options.txt", which isn't in the latest subversion
checkout. (r28304 2009-01-11 14:48:51). Is there any reason I should
avoid vqscale=1 when encoding mpeg2video ? Also, where did
libavc-options.txt go? The title makes it seem like a useful read.

3. I still don't understand the "pullup" filter 100%. I searched the
archives and found a couple of stuff. I just want to make this clear
to myself... is "Skipping frame!" a good message with pullup? And
similarly... is "Duplicate frame" a bad message? I'm tired of looking
for panning scenes manually to see if the video pans unevenly.

4. What is a good PSNR value? I'm hitting "All" PSNR average == 47.50
with my current encoder settings... (and a minimum of ~42.5 PSNR...
found with awk -F, ' BEGIN {min = 100000} // {if(min > $7) min = $7;}
END {print min}' psnr_173418.log) and I'm not really sure if thats a
good or a bad number. All I know is that "bigger is better". Currently
I'm only getting ~350 MB per 30 minute episode, so if there are other
options that increase PSNR at the cost of space, I'm willing to go for
it. (aside from switching to lossless, of course).

5. Finally, is there any use of 2-pass encoding if I don't care about
file size? I'm using constant quantizer (vqscale=1 specifically)...
and I have the impression from the manual that 2-pass encoding will
only improve the size of my file... and not the quality.

Thank you for reading this post. Finally, here is my current settings:

mencoder Cry9.mpg -o Cry9_2.mpg -vf pullup,softskip -oac copy -ovc
lavc -lavcopts vcodec=mpeg2video:vqscale=1:last_pred=3:mbd=2:dia=4:trell:psnr
-of mpeg -ofps 24000/1001

Options I'm experimenting with:
cmp=2 is ~2 to ~3x slower, with only PSNR value of ~.1 better.
preme=2:predia=4 improves PSNR by ~.01.

-- Percival Ross