[MPlayer-users] Linux port of Donald Graft's "Decomb" deinterlacer for mencoder?
John Evans
jevans64 at knology.net
Sat Jan 18 06:34:26 CET 2003
Hi All,
I am looking for a good postprocessing utility for mencoder. My encodes
look great using the Decomb postprocessing to deinterlace certain TV
series on the Windows side but look like **** when I use mencoder pp=lb
on the Linux side. I have proof of this in the form of sample encodes
using both Virtual Dub and Mencoder. I will send the samples to anyone
who requests them since they are too large ( ~1 MB each ) to post here.
The problem is that there are 2 good frames followed by 2 blurred frames
( looks like 2 different fields that are combined into 1 frame ). It can
be shown best by my samples but here is a diagram of what is happening.
frame 1 >> | 1 | 2 | 3 | 3 | 4 | 5 | 6 | 6 | 7 |
frame 2 >> | 1 | 1 | 2 | 3 | 4 | 4 | 5 | 6 | 7 |
Each number represents a different field. Numbers between the pipe
symbol represent a frame. Numbers that are the same between frames
represent the same field data as the previous / next frame. I have tried
the "il=" command to attempt to swap field order with the same results.
I have also tried 29.97. I just get more blended frames. Using 27.97
with Decomb yields all fields matched with a duplicate frame every 5
frames. So 23.976 is correct in this case to get fluid motion.
Sample script >>
mencoder -v -mc 0 -o UFO_109_1780_23.avi -oac copy -aid 128 -ovc lavc
-lavopts vcodec-mpeg4:vbitrate=1780:vhq:vqmin=2:vqmax=3:keying=500 -ofps
23.976 -ni -vop crop=632:480:3:0,scale=640:480,pp=lb -sws 2 ./UFO109.vob
-- I did the entire series using the same script ( except for names,
crop, etc. ). Episode 109 was the only one with combined fields. I used
a similar script for Capatin Scarlet series which resulted in 2 episodes
with combed frames and all episodes of Thunderbirds were combed.
The source for Decomb can be found in the links at
www.shelob.mordor.net/dgraft/news.html
Thanks in advance for any solutions,
John
More information about the MPlayer-users
mailing list