[FFmpeg-user] How can I use the fieldmatch filter to detelecine?

[L. Lee]

On 4/17/13 4:57 PM, "Andy Furniss" <andyqos at ukfsn.org> wrote:

>L. Lee wrote:
>> I've developed a highly refined method for applying detelecine using
>> MEncoder's pullup filter methods. I get excellent results using that
>> method for the examples I'm including in this message. However, for
>> reasons that are probably obvious, I'm now trying to get comparable
>> results using ffmpeg.
>>
>> How can I use the fieldmatch filter to detelecine these excerpts to get
>> better results using ffmpeg? This source (for both source 1 and source
>>2)
>> is 720P. I don't see any difference in my results whether or not I use
>> decimate in the filtergraph as shown in the 8.22.2 Examples
>> (http://ffmpeg.org/ffmpeg-filters.html#toc-Examples-3). I've also
>>included
>> a slightly longer mpg-muxed version (source 2) which concludes with the
>> source 1 excerpt. I get different results when encoding the longer
>> mpg-muxed excerpt (detelecine seems correct, however some areas are
>> pixilated, but only when encoding the longer excerpt) Although the
>> included filtergraph for my source 2 command uses the shorter of the two
>> examples provided in the 8.22.2 section of the documentation, I've tried
>> both and can't observe any difference in the output.
>>
>> Here's my source 1:
>>
>> https://dl.dropboxusercontent.com/u/42718751/Blockyclip.zip
>>
>> Here's my output 1:
>>
>> https://dl.dropboxusercontent.com/u/42718751/Blockyclip_output.zip
>
>This looks wrong - 12fps with frames duped.
>
>_______________________________________________
>ffmpeg-user mailing list
>ffmpeg-user at ffmpeg.org
>http://ffmpeg.org/mailman/listinfo/ffmpeg-user

Thanks, also, for your reply, Andy. Yes, I agree, it was wrong. I'm making
progress now. The 1080i conversion is working OK with regard to fps and
correct detelecine, and by using "interlace,fieldmatch" in the
filtergraph, that's also true for the 720p. But I'm still having trouble
with blockiness (pixelization) as observable in the sample provided, but
now I'm trying to determine whether my method of building ffmpeg needs
correcting (see my last post). Thanks again.

Laine Lee