[FFmpeg-user] minterpolate problem

[Paul B Mahol]

On Thu, Jan 28, 2021 at 7:04 AM Mark Filipak (ffmpeg) <markfilipak at bog.us>
wrote:

> On 01/27/2021 05:47 PM, Carl Eugen Hoyos wrote:
> > Am Mi., 27. Jan. 2021 um 02:03 Uhr schrieb Mark Filipak (ffmpeg)
> > <markfilipak at bog.us>:
> >
> >> ffmpeg -i SOURCE -map 0 -filter_complex "split[1][2], \
> >> [1]telecine=pattern=55, select=not(eq(mod(n\,5)\,2))[3], \
> >>
> [2]minterpolate=fps=60000/1001:mi_mode=mci=mc_mode=obmc:scd=fdiff:scd_threshold=10,
> \
> >> select=eq(mod(n\,5)\,2)[4], \
> >> [3][4]interleave" \
> >> -codec:v libx265 -x265-params "crf=16:qcomp=0.60" -codec:a copy
> -codec:s copy TARGET
> >
> > Since others may read this:
> > The command line makes little sense, you should either use
> > telecine or minterpolate.
>
> Carl Eugen, you surprise me. I'm sure you see what I'm trying to do, so
> why the angst?
>
> First, let me admit that since I started this thread (following weeks of
> experimentation) I decided
> to simply try this: 'minterpolate=fps=60000/1001' to see what it would do.
> I was amazed.
>
> If minterpolate is Paul's work, then Paul is a f'ing genius.
>
> I've since spent a couple of days playing with minterpolate and have found
> that this:
>
> 'minterpolate=fps=60000/1001:mi_mode=mci:mc_mode=obmc:scd=fdiff:scd_threshold=10:vsbmc=1:search_param=20'
> is astounding.
>
> Nonetheless, I'd still like to generate a 60 FPS transcode that is
> visually equivalent (as closely
> as possible) to 24 pictures/second by replacing the combed frame (i.e.
> n%5==2) by an interpolated
> frame -- hence the filter graph you loathe.
>
> For that n%5==2 frame, here's what I sought (and did not find):
>
> Sobel of line y:   ______/\_____________/\_________
> Sobel of line y+1: __________/\_____________/\_____
> Desired output:
>           line y:   ________/\_____________/\_______
>           line y+1: ________/\_____________/\_______
> I could find this:
>           line y:   ______/\_____________/\_________
>           line y+1: ______/\_____________/\_________ (aligned to top line)
> and I could find this:
>           line y:   __________/\_____________/\_____ (aligned to bottom
> line)
>           line y+1: __________/\_____________/\_____
> but I could not find this:
>           line y:   ________/\_____________/\_______ (aligned to mean)
>           line y+1: ________/\_____________/\_______ (aligned to mean)
>
>
Please cite reference/paper that explains this technique in more detail.



> One reason I might -- MIGHT -- prefer telecine55 is that the original
> frames are recoverable from
> the transcodes -- but admittedly, that's not much of a reason, is it?
> Another reason is that the
> transcode is much, much quicker than 'minterpolate=fps=60000/1001' and
> somewhat smaller.
>
> There you have it, Carl Eugen. Please stop trying to make the world think
> I'm feeble minded.  :-)
> Okay?
>
> Regards,
> Mark.
>
> --
> Someone's sneaking in and turning up the range so that my food burns.
> I'm sure of it.
> And the older I get, the more sure of it I become.
> _______________________________________________
> ffmpeg-user mailing list
> ffmpeg-user at ffmpeg.org
> https://ffmpeg.org/mailman/listinfo/ffmpeg-user
>
> To unsubscribe, visit link above, or email
> ffmpeg-user-request at ffmpeg.org with subject "unsubscribe".