[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: 1080i color space?

Rob asks what the difference between 601 and 1080i color is.

There are two differences, one of which I can simulate on your computer if
you want.

First, they use slightly different primaries on the CIE x-y plane -- that
graph that looks like a shark fin.

Second, they use different luminance equations, and consequently, different
difference matrices in converting between RGB color and luminance/color
(Actually, from gamma corrected RGB to gamma corrected Y Cr Cb.  Poynton
using luminance for the linear version and luma for its gamma corrected

The proposed color standard for 1080i and 720p is called ITU Rec 709, while
ITU Rec 601
is proposed for all the 480 line formats.  There's little if any 709
implemented yet.

There are two guys on the planet who really grok this stuff, LeRoy DeMarsh,
and Charles
Poynton.  I've talked to both, and I can summarize their positions on this
simply:  They both say that 709 is a mistake.  We'd do better to use 601
for everything.

The little program takes any 24 bit .BMP file, and spits out two modified
One showing what happens if you encode as 601 and decode as 709, and the
other what
happens if you encode as 709 and decode as 601.

The biggest numerical differences are in green.  Some pictures from last
of kids in costume show this, since they contain saturated green and
orange.  Nature
shots with lots of grass and trees don't show as much of a problem.  Faces,
the numerical differences are fairly small, do have more of a problem,
because we're
quite sensitive to errors in that specific subject matter.

I sent the sample pictures and the source and executable to Craig
Birkmaier, who
posted them on his web site.  They're probably still there, www.pcube.com.

If not, I can send them to you (Rob) if you want.

The hell of it is, the errors are of the worst possible magnitude:  a
little less than
the difference between TV sets at Circuit City.  They're big enough to
aggrivate the
artists, but small enough that the bean counters will resist paying to fix

-- J.S.

Thanks to Aaton and Jean-Pierre Beauviala for support in 1998.
No product marketing allowed on the main TIG.  Contact rob at alegria.com
1009 subscribers in 38 countries on Wed Nov  4 20:53:37 CST 1998 
subscribe/unsubscribe with that Subject: to telecine-request at alegria.com
complete information on the TIG website http://www.alegria.com/tig3/