[MPlayer-users] new inverse telecine filter!

[D Richard Felker III]

Hey, I've just committed a new filter for inverse telecine, much
better than the old one. Invoke it with -vf ivtc, and an optional
argument 1 if you want frame dropping (for use with mencoder).
Hilights include:

1. Pretty much entirely stateless, so it doesn't get stuck on an
   incorrect pattern like the old detc filter.

2. Looks at the next frame when deciding what to do with a frame --
   this gives it the ability to properly restore interlaced frames
   that come right after a scene change.

3. Works on all the truely-hard-telecined material I've tested,
   including stuff on MPHQ (Nausicaa.mpg, flcl-detc.avi,
   flcl-explosion-detc.avi), SE:Lain DVDs, Dolby demo VOBs, and files
   generated by -vf telecine.

4. New MMX code for the interlacing metrics makes it a lot faster.

5. New interlacing metrics more accurately identify the pulldown
   frames. Unlike the old detc filter which can lock onto a pattern,
   ivtc has to identify *each* interlaced frame, and from my
   experience the new code can actually do that.

There are a few known problems, however:

1. No way to deal with multiple out-of-sync telecined subpictures in
   the same picture. AFAIK there is no inverse-telecine engine to
   handle this situation yet, although in theory there are some
   possible tricks to do it.

2. Files that use bff (bottom field first) telecine have to be flipped
   before and after running the ivtc filter (foo-sample.mpeg).

3. What's worse, files that mix tff and bff using weird combinations
   of mpeg2 header flags cannot be fixed at this time due to mplayer's
   architecture. One example is samurix_detc.avi on MPHQ.

Please experiment and report any problems found, preferably in replies
to this message so that the thread stays together! Happy ripping!

Rich