[Tig] questions about gamma correction in video
Thu Dec 8 20:47:55 GMT 2005
Thanks to oktobor for supporting the TIG
I'm doing a little research into this and finding conflicting answers.
1a- What is assumed to be the transfer function of a CRT display? (Or what
is it supposed to be?)
Charles Poynton indicates on his website that it is 2.5, not 2.2 as other
(reputable?) sources say.
1b- What is the actual transfer function / "gamma" of a reference-grade
According to the following PDF, it's about ~2.38 for a BVM-D32 (not sure if
the figure is for an aged one or not). *The 2.38 figure is just my GUESS
from looking at the excel graph in there, NOT an actual figure presented in
This display technology shootout cites 2.2 gamma as ideal, and says the Sony
PVM-20L5 has a measured gamma of 2.2. It also says that Ikegami monitors
have a gamma of 2.2 (according to their director of technology). The author
runs Display Mate and has a phD, so I presume he is fairly reputable.
So which figures are correct?
2a- Poynton suggests in his gamma FAQ that an end-to-end power function can
be used to compensate for a dim surround. From messing around with
Photoshop, I find this to be very objectionable. I based this off the
original PSD used to generate the images at
I played around with the "real world" image with the building on the
articulated background. It's similar to the classic simultaneous contrast
demonstration, except the inner square is replaced with a real world image
and the outer part has checkerboarding. The inner images are identical
except for the two different surrounds, one bright one dark.
As soon as you mess with the gamma (in Photoshop via the Levels filter), it
causes saturation shifts- the image overall looks like it has greater or
less saturation, depending on which direction you go.
If I add an adjustment layer to make everything black and white, I find that
I want to go the other direction with the levels/gamma to make the images
look closer. These corrections were done in Photoshop's "RGB" space (which
maybe should be called R'G'B'?).
If I switch into LAB mode and affect "gamma" that way, then the two images
do look more similar without very obvious hue or saturation shifts. To me,
this is less objectionable. But I don't think any displays process color in
LAB or any other color space with an independent luminance component.
Basically what I'm wondering is, am I missing something here? Is it just me
or is an end-to-end power function a good idea?
2b- In the human perception quirks article (see above), I suggest that the
simultaneous contrast is very weak or non-existent for real world images.
This would suggest that no end-to-end power function or correction is
necessary at all.
*Looking at the image again, I do find that the real world image in the dark
surround has a lot more "pop" and contrast and saturation than the image in
the bright surround. My definition of contrast: it's subjectively how much
an image pops, which is not quantifiable. Poynton may be using a different
definition- he suggests that a dark surround causes LESS contrast, which is
the opposite of what I'm seeing.
Tig list - http://tig.oktobor.com/mailman/listinfo/tig
TIG wiki: http://tig.colorist.org/wiki
More information about the Tig