24 Aug
2015
24 Aug
'15
7:55 a.m.
Am 24.08.2015 um 00:33 schrieb Carl Hetherington via DCPomatic:
Hi Jim,
Apologies for the delay in looking at this. I will investigate what
FFmpeg (the video decoding library used by DCP-o-matic) does with
respect to video level headroom.
As 16-235 is intrinsic with Rec. 709, I would expect 16-235 to be stretched into DCI 12Bit
xyz. Best would be to create some test images and then look into the resulting xyz J2Cs to
find out. It's actually hard to detect on screen, both in shadows and highlights. One
could probably measure it.
There could be issues with x.v. color encoded files - some cameras have options to use
that extended color space. There are markers in common video formats which *SHOULD* allow
these variants to be detected automatically. Wether that is handled properly by FFMPEG, I
don't know.
Yes, this is all painfully complex and unsatisfactory, also that there is no single Rec.
709 file format, but so many different codecs and container types using Rec. 709, and you
can never be absolutely sure on a formal level. It may work the right way with one camera
source file format, but not with another. And then we did not even touch deep color Rec.
709 at 10 or 12 bits/pixel...
Would be nice to have some sort of vectorscope or waveform monitor in DCP-o-matic for the
outgoing xyz signal levels - in a similar way it exists for audio.
- Carsten