in the broadcastworld yuv-data is mostly 10bit yuv
(digital betacam, d5...) would be nice to have a 10bit
yuy2 and maybe a 4:4:4 colorspace in avisynth.
A planar YUV 4:4:4 is not a big deal in AviSynth, as most
code from YV12 could be used.
A 10 bit per channel format would not be practial as a
planar format, as there will be used 16 bits per pixel
element in memory anyway. It wouldn't make much sense as an
interleaved format either, as it is very nasty having to
unpack 10 bits at the time.
But the most useable format would probably be a planar 4:4:4
15 bpc format - this would provide more than enough
precision, and still be able to run quite fast. All filters
would of course have to be rewritten almost completely to
support this.
Actually it will be 16bps, but with the most significant
byte being unused, to avoid singned/unsigned problems when
dealing with shorts. This would allow YUV values from 0 to
32767.
A 16 bpc will take as much time to implement as the initial
YV12 took. A planar 4:4:4 will take less than half - down to
a fifth or so.
First we need a non-beta 2.5 though.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
As most algorithms (ie resizers, smoothers, etc) are
actually intended to work with a linear scale of lightness
it might be an idea to unwind Gamma going into 15 bits and
rewind it going back to 8 bits. There is certainly more
than enough precision in 15 bits to do this.
So the Avisynth standard for 15bpp data is that
coresponding light levels are linear! And that 8 bit/10 bit
data remains Gamma corrected as is now.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Logged In: YES
user_id=578300
A planar YUV 4:4:4 is not a big deal in AviSynth, as most
code from YV12 could be used.
A 10 bit per channel format would not be practial as a
planar format, as there will be used 16 bits per pixel
element in memory anyway. It wouldn't make much sense as an
interleaved format either, as it is very nasty having to
unpack 10 bits at the time.
But the most useable format would probably be a planar 4:4:4
15 bpc format - this would provide more than enough
precision, and still be able to run quite fast. All filters
would of course have to be rewritten almost completely to
support this.
Actually it will be 16bps, but with the most significant
byte being unused, to avoid singned/unsigned problems when
dealing with shorts. This would allow YUV values from 0 to
32767.
A 16 bpc will take as much time to implement as the initial
YV12 took. A planar 4:4:4 will take less than half - down to
a fifth or so.
First we need a non-beta 2.5 though.
Logged In: YES
user_id=578300
Priority is fairly low, as new colorspaces are not top-priority.
Logged In: YES
user_id=673887
As most algorithms (ie resizers, smoothers, etc) are
actually intended to work with a linear scale of lightness
it might be an idea to unwind Gamma going into 15 bits and
rewind it going back to 8 bits. There is certainly more
than enough precision in 15 bits to do this.
So the Avisynth standard for 15bpp data is that
coresponding light levels are linear! And that 8 bit/10 bit
data remains Gamma corrected as is now.