[FFmpeg-user] Questions about w3fdif deinterlacing

[Seth Parker]

I just had a project come up where I needed to batch encode a video archive
to H.264. the problem was that the archive contains both progressive and
interlaced content and the interlaced content would need to be
deinterlaced. From my Googling, it seemed that there isn't an easy,
programmatic way to only deinterlace the interlaced streams. Most
suggestions said I had to visually inspect the streams and selectively
encode based on that.

Then I came across the recently released w3fdif filter which has an
"interlaced" option that the ffmpeg documentation describes as, "Only
deinterlace frames marked as interlaced." However, I can't find any more
documentation on how exactly that works and whether or not its interlaced
detection is good enough for the sort of batch processing I need to do. Can
anyone elaborate on this filter and its usage?

As a side note, when I first started looking into w3fdif, I thought that I
was supposed to use idet in the filter graph prior to w3fdif so that it
knew what frames were and were not interlaced. This didn't seem to make a
difference on my 29.97fps content, but I found a really strange issue with
59.94fps (DVCPro HD) footage. Using idet before w3fdif created a 59.94,
deinterlaced stream as expected. However, if I removed idet, the output
became choppy and unviewable. This at first confirmed my understanding of
how idet/w3fdif worked, but the more I read into idet, the more I thought I
had it all wrong. Now I'm just confused. So if anyone can speak to idet
usage and perhaps help me understand what happened with my 59.94fps
footage, I would greatly appreciate that as well.

*Seth Parker*
Video Editor | The University of Kentucky Vis Center
www.csethparker.com | www.vis.uky.edu