[FFmpeg-user] Another odd colourspace issue

[Mark Himsley]

On 10/07/12 12:34, Tim Nicholson wrote:
> I have a programme that generates packed yuv test signals.
>
> Using the programme I generate 100/0/100/0 colour bars and greyscale,
> which I then convert to mov's using a command of the form:-
>
> ffmpeg -pix_fmt uyvy422 -s $size -i $in_file -c:v copy -vtag 2uvy $out_file
>
> I have created both SD (720x576) and HD (1920x1080) frame size files
> using both 601 and 709 coefficients, as described at
> http://avisynth.org/mediawiki/ColorBars_theory and verified by my own
> calculations from first principles.
>
> When I inspect the files on both FCP and Kdenlive waveform monitors
> there is am marked difference in levels between the HD and SD versions
> using the same colourspace. The 709HD and 601SD look correct, but the
> 709SD and 601HD look wrong. Based upon the errors in the green channel
> it is clear that different coefficients are being used to decode than
> were used to generate the original material, and these coefficients seem
> to be being auto selected on the basis of the frame size.
>
> The question is why? and what is doing the selection?
> Is it both Kdenlive and FCP deciding on the basis of framesize, or is
> ffmpeg adding some kind of colourspace marker to the file when it
> rewraps to mov?

Hey Tim,

I believe that the colour space is generally decided solely on the frame 
size. I believe from my experience that is the way FCP does it. I'd love 
to be proved wrong - because I think that is cack.

-- 
Mark