[FFmpeg-user] Inverse Telecine with fieldmatch filter

L. Lee llee040 at sbcglobal.net
Wed Jun 26 21:46:32 CEST 2013


On 6/26/13 11:11 AM, "Carl Eugen Hoyos" <cehoyos at ag.or.at> wrote:

>L. Lee <llee040 <at> sbcglobal.net> writes:
>
>> >> For 720p I used your suggestion to interlace and deinterlace:
>> >
>> >(I should add that I meanwhile realized it may be
>> >faster to use fps=30000/1001 instead of interlace)
>> 
>> You're not suggesting that the output file should be
>> 30 fps, are you?
>
>(skip this)
>
>No, I am suggesting that instead of using the interlace
>filter (that needs not much but some computation power)
>you can use the fps filter that does the same for this
>use-case (and should need no computation power at all).
>
>If you cannot reproduce a speed gain, it may make no
>difference.
>
>(until here)
>
>What I probably meant (iirc) was that you could try to
>remove both the interlace and the fieldmatch filter and
>instead use fps (as above) and decimate.
>If you trust the television provider to have done their
>work correctly, the output should be identical but
>much less cpu power will be needed.
>
>(Remember that your input is not strictly telecined but
>telecined and deinterlaced, it may also be "telecined"
>by simply repeatings frames - instead of fields - which
>would be simpler for the provider and would not need
>fieldmatch at all.)


I think my problem with the concept is that I'm used to specifying both
input frame rate and output frame rate in MEncoder, and to my way of
thinking, your reference to fps=30000/1001 must fit into one or the other
of those categories. No matter what pullup conversion method I use in
MEncoder, if I don't specify -fps=30000/1001 and -ofps=24000/1001, the
inverse telecine operation won't be successful. Does fps=30000/1001 at the
beginning of the filtergraph as included in your previous offering tell
FFmpeg to read the file as if it was 29.97 fps as the -fps=30000/1001
option does in MEncoder?

Also, although in my previous message I stated that I wasn't any longer
having issues with the transcoding itself, a new one has arisen. As I
already said, the example for which I originally included my 1080i
transcoding command and terminal output was entirely successful. However,
I have since tried the operation on another source stream that has the
same specifications almost to the letter, but the results leave much to be
desired (jerkiness, interlaced frames, and skipped frames abound). Please
keep in mind that my MEncoder results are flawless for both of these
streams, using precisely the same command line except for file references.
If I provide links to two source excerpts for download, the FFmpeg command
and output video files as well as links to output video files from
MEncoder for download, is there a chance that you could help me find the
cause of the issue and fix it?

Again, your help is greatly appreciated. Thanks again.

Laine Lee




More information about the ffmpeg-user mailing list