[FFmpeg-user] Confusion about fieldmatch, decimate and framerate (now with yadif)
Nicholas Robbins
nickrobbins at yahoo.com
Thu Dec 19 03:02:27 CET 2013
>> fps=fps=30000/1001,fieldmatch,decimate,yadif (which will yadif
>
> I would surely like to get a sample. I shouldn’t need more than 20-30
> seconds to see what results I’m able to produce. You might be able to
> email me such an excerpt. I don’t see how any conversion to deinterlaced
> 30 fps would result in optimal output. Nearly any smooth horizontal
> panning is like getting slapped in the face 6 times per second. If output
> is deinterlaced, all 4th and 5th frames are duplicates. If not
> deinterlaced, those frames are interlaced complements which still take
> twice as long as a normal frame to play. Thanks.
>
>
> Laine Lee
>
I will see about sending you a snippet offline. I can't seem to edit the file as a vob without starting at the beginning and there aren't really good problems at that point. So I've remuxed it into a mkv using mkvmerge.
Also, I don't think you are quite understanding the filters. fps=30000/1001 is actually just relabeling things. If you've been following the thread, the video file actually has this man frames, but the container things is has 60000/1001 fps. So that first filter just labels the file correctly. The second filter should replace the 30 fps telecined with ABCDD non-telecinded originals. Decimate will then remove every fifth frame. At least that is how it is supposed to work. However, it doesn't always work, some frames sometimes are missed, sometimes due to editing after telecining for instance. These frames are supposed to me marked as interlaced by fieldmatch and then yadif can de-comb them, which isn't ideal, but the information to do the ideal thing is on the cutting room floor presumably.
However, yadif=deint=interlaced isn't noticing any interlaced frames. So if I want to remove the combing, I have to yadif everything.
In any case, I'll send you the file and you can play with it.
-Nick
More information about the ffmpeg-user
mailing list