[FFmpeg-user] Confusion about fieldmatch, decimate and framerate (now with yadif)

[Carl Eugen Hoyos]

Nicholas Robbins <nickrobbins <at> yahoo.com> writes:

> Done. Uploaded as robbins-recut.tgz.

Sorry, I fear that for different reasons I will not be 
able to help you, still some comments / rants that may 
or may not help:

I live in PAL-land and I only know telecined content 
from the (old) MEncoder documentation and a few 
samples on samples.ffmpeg.org
For a long time, people on mencoder-user explained 
that they cannot use FFmpeg because of the missing 
inverse telecine filters. The mp=pullup filter did 
never work, so we were hoping for somebody to port 
it to a native filter. At the same time, different 
people repeatedly reported that it would not make 
sense to port pullup because it is outdated and 
better filters exist. I don't know how often I 
asked for samples, but no samples were ever offered, 
I was just accused of being an MEncoder-lover.
(I may be but I never understood how this was 
related.)
Finally, fieldmatch was written, a filter that is 
said to be far superior over pullup and that was 
supposed to fix all problems. Apart from the fact 
that the decimate filter (which is needed for 
fieldmatch) has problems of which only some were 
fixed (by me!) the filter was shown to be very 
slow.
Finally, pullup was ported and now, you already 
uploaded the second sample that works with 
pullup but not fieldmatch. (I think it was you.)

The sample you uploaded does not show a 2:3 pattern 
(and no 2:3:3:2 pattern) but a 3:3 pattern.
I have no idea how this was made or how it is 
supposed to work. (Did you make the sample?)
fieldmatch does not seem to like it, it shows 
surprising output on the console and I have no 
idea if the original avisynth (or whatever) 
filter has the same deficiency or not. (Can you 
test?)

All this said, the sample looks interesting, if 
you could explain where it comes from, perhaps 
we can find out what has to be done.

Sorry for the rant, Carl Eugen