[FFmpeg-user] Objectively best deinterlacer?

[Robert Krüger]

On Fri, Jan 24, 2014 at 5:24 AM, Carl Eugen Hoyos <cehoyos at ag.or.at> wrote:
> Robert Krüger <krueger <at> lesspain.de> writes:
>
>> However, in the case of 12-bit Prores you will have
>> your input reduced to 10 bit but that is a limitation
>> of ffmpeg's Prores implementation (decoder(s) and
>> encoder(s)) rather than yadif.
>
> I thought this may at least be a limitation of the
> QuickTime documentation, is there any hard evidence
> that this isn't the case?

Hard evidence in the sense of analyzing the numbers of the pixel
values in a workflow is difficult to get as very few commercial
software components support a 12bit workflow (even Apple's). I had
tried to obtain that a few months ago but then stopped as it suddenly
wasn't important for my project anymore.

Evidence so far:

- Apple Whitepaper
- Officially licensed products (hardware recorders) that claim to output 12 bit
- Ronald's explanations about the code decoding 12bit coefficients and
then scaling them down (IRC)