Hello,

I am trying to efficiently extract motion vectors from h264 live
stream for analysis and motion detection. Is there any way to parse
motion vectors without actually decoding the whole frame? I am aware
of 'extract_mvs.c' code, but when I was experimenting with it, the
performance was pretty much the same as decoding h264 to rawvideo (I
think it completely decodes every frame) so right now I see no benefit
in performing analysis in compressed domain vs. pixel domain. In
theory it should be possible and performance gain could be impressive.

Thank you for your time

Best Regards
Štěpán Šigut
_______________________________________________
Libav-user mailing list
[email protected]
http://ffmpeg.org/mailman/listinfo/libav-user

Reply via email to