[Openicc] Linux CM ideology, was: meta data in test chart

[Graeme Gill]

edmund ronald wrote:
>   I think the TV movie guys and camera guys are going to end up
> figuring out a model which works for them, just to get things working
> reasonably well - customers who walk into a TV store expect to compare
> TVs, cinematographers expect their theater and TV audience to share an
> experience etc - I think at some point these people are going to

The evidence is against this. The majority of customers (judging by
the "features" in modern TV sets), care little about color, as long
as it impresses them. They aren't interested in accurate or realistic
color, just color with "impact". Hence the plethora of dynamic features
that manipulate the image, from frame up-interpolators (something that
should be unnecessary if the original images are sampled properly to
include motion blur), to edge enhancers, color enhancers, and
color and range "enhancers" and filters of all descriptions.

To get what I would call realistic pictures out of a modern TV
requires turning off upward of half a dozen of these "features".

Wide gamut ? Much like "D3" (read stereo vision), the problem
will be "solved" by taking normal gamut signals, and "enhancing"
them (there have already been papers at the CIC on this, and many
more in less color savvy technical publications). No need to
accurately convey a wider gamut, simply compress wide gamut originals
into normal gamuts, then "enhance" at the TV end, so that
_every_ picture has eye popping color! That's how to sell TV sets,
make sure that your set is more eye-popping than the one next to it!

[ I understand that some sets have a "store" setting to
  ramp up the picture, and "normal" setting for actual viewing of
  progran. ]

Graeme Gill.