[FFmpeg-user] nvenc h264 encoding

Nicolas George george at nsup.org
Thu Oct 13 18:05:32 EEST 2016

Le duodi 22 vendémiaire, an CCXXV, Sven C. Dack a écrit :
> If I'm not mistaken then yuv444p is linked to 10-bit depth encoding

Not necessarily. yuv444 is opposed to yuv420: with 420, pixels are grouped
as 2×2 squares, where all four pixels have their brightness ("y" =
"luminance") encoded, while the color ("u"/"v" = "chrominance") is the same
for all four; with 444, pixels have all three components coded individually.
That means that in 444, each pixels use three components, while in 420 they
only use 1.5 on average, half as much. This makes sense because the eye is
more sensitive to brightness changes than color changes.

10-bits depth is opposed to 8-bits depth, it measures the accuracy for each

All four combinations are possible: yuv420p8, yuv420p10, yuv444p8,
yuv444p10. The 8 is usually not printed because it was for a long time the
only possible choice and is still the default. Of course, because of
compatibility problems, once you decide to ditch yuv420p8, you might as well
go all the way to yuv444p10, therefore the two intermediate ones are more
rarely seen. But yuv444p8 makes sense for screencasts, to avoid color

The p means that libavcodec stores the three components in three separate
memory areas rather than interleaved all together like rgb24.


  Nicolas George
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 801 bytes
Desc: Digital signature
URL: <http://ffmpeg.org/pipermail/ffmpeg-user/attachments/20161013/a728861a/attachment.sig>

More information about the ffmpeg-user mailing list