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

Nicholas Robbins nickrobbins at yahoo.com
Sun Nov 10 17:20:59 CET 2013

On Sat, 11/9/13, Carl Eugen Hoyos <cehoyos at ag.or.at> wrote:

 Subject: Re: [FFmpeg-user] Confusion about fieldmatch, decimate and framerate (now with yadif)
 To: ffmpeg-user at ffmpeg.org
 Date: Saturday, November 9, 2013, 11:52 PM
 >Nicholas Robbins <nickrobbins <at> yahoo.com> writes:
> > Ok, on my current sample. I checked the length and 
>  > framecount using ffmpeg, getting: Frames: 169698
>  > Duration: 1:34:19.29 for a frame rate of 29.97.
> I apparently miscalculated, sorry.
 No problem. Good to know that this is the right way to determine framerate.

> > In the encode, I get this quite a bit: 
> > "Frame #328307 at 10954.5 is still interlaced" So I think, 
> > I should add> in yadif=deint=interlaced as per 
>  > http://ffmpeg.org/ffmpeg-filters.html#Examples-37 between 
>  > the fieldmatch and the decimate.
> Sounds correct and reasonable to me.

>  Did you extract the offending frame as fieldmatch outputs 
> it and compared it with the same frame as output by 
> fieldmatch,yadif=deint=interlaced ?
> Every other test is meaningless.

I have done so now. They appear almost identical (as far as I can tell using gimp). The commands I used were:

ffmpeg -i start.mkv -map 0 -c copy -c:v libx264 -vf "fps=30000/1001,fieldmatch,yadif=deint=interlace" -crf 17 -preset veryfast fieldmatch-yadif.mkv

ffmpeg -i start.mkv -map 0 -c copy -c:v libx264 -vf "fps=30000/1001,fieldmatch" -crf 17 -preset veryfast fieldmatch.mkv

both of which produced:

Frame #8158 at 272.239 is still interlaced time=00:04:32.25 bitrate=2560.3kbits/s
[Parsed_fieldmatch_1 @ 0x6c7357b3e0] Frame #8159 at 272.272 is still interlaced
[Parsed_fieldmatch_1 @ 0x6c7357b3e0] Frame #8160 at 272.305 is still interlaced

among many other similar lines in their output. I picked this run because it was the longest. Most were 2 frames.

Then I used 

ffmpeg -i fieldmatch-yadif.mkv -frames 8170 /var/srv/media/videos/Testing/Postal/jpegs/fieldmatch-yadif-%d.bmp (and without the -yadif) 

To get the relevant frames.

> ( This tests something else which is 100% unrelated 
> (and which you should really ignore / forget if 
> you are interested in inverse telecine, there is 
> no real world stream that sets this the way you 
> test it.) 

Ok, I'll ignore that flag.
> > I guess fieldmatch doesn't mark the frames it can't 
> > deal with. It just leaves them where they are.
>  This would be a serious bug, if you can actually reproduce
>  it, please upload the five-ten relevant frames.
Done. I've put them in robbins-fieldmatch-yadif.tgz in incoming.

> Carl Eugen

 ffmpeg-user mailing list
 ffmpeg-user at ffmpeg.org

More information about the ffmpeg-user mailing list