[MPlayer-users] Inverse Telecine and JPEG Extraction

[James Board]

>Please report on ffmpeg-user if the following does not work:

>ffmpeg -i IN.avi -vf pullup,fps=24000/1001,split[a][b] \
>-vc ffvhuff -map [a] OUT.avi -map [b] OUT%4d.png
>
>You can also use fieldmatch, it may produce better results 
>but is slower.

I was able to get the -vf fieldmatch,decimate working with ffmpeg.
However, it is 3x slower as you mentioned. I don't want to
complain about the speed on ffmpeg-users, unless you think
there's a chance they can make it run faster.

I'm considering splitting the input file into multiple
segments, then doing the IVTC step with ffmpeg in parallel
on different cores, if that will improve the speed.  However,
when I seek to the different locations in the input file with -ss/-to
it seems to walk through every frame, so there's no speedup.
I probably have to stick with mencoder (which I'm happy with).

>Considering that you in the past have requested "frame 
>accuracy" (maybe I misunderstood): I believe that it is an 
>unavoidable characteristic of an inverse telecine filter 
>(used on real-world input!) that it is not "frame-accurate" 
>in the sense that I believe you requested it in past mails:

Frame-innaccurate for only the Inverse Telecine step is okay for me
since I do that before editing.

Thank you.