On Oct 1, 2009, at 9:50 AM, James Paige wrote:
On Thu, Oct 01, 2009 at 09:34:35AM -0400,
[email protected] wrote:
Sounds lik yet another gimmick to get uneducated folk to buy
another TV
LOL, I like this one.
But I'm not sure I understand your statement on NTSC.
This is what my research subject is all about. My boss asked me to
optimize my hardware design (0.35u CMOS image sensor) to fit eyes and
equipement limitations. But I need first to create RGB101010
software
to see if DeepColor makes sense or not, before optimizing CMOS chips.
Since I like Python and I have Pygame experiences, I wanted to do
that
software with Pygames.
Thanks
Pierre
Regarding the limitations on the human eye: I am no expert on this,
but
from what I have read, the average human eye can distinguish somewhere
from 7 to 10 million different colors.
RGB888 is 24 bit color, which is enough for over 16 million colors.
So by that measure, average human eyes should not be able to tell the
difference between RGB888 and RGB101010
Of course, I am no expert on this, and I know that the way the rods
and
cones in an eye encode color is very different from the way digital
color is encoded in pixels, so there may possibly be other reasons why
RGB888 might be less that mathematically perfect for human color
vision.
I would be interested to hear from someone who IS an expert. Do we
have
any Opticians on-list? :)
The advantage with having more colors is more about dynamic range than
how many colors can be perceived. 8 bits per color effectively gives
you only 256 different color intensities, whereas 10 bits gives you 4x
more, and floating point rgb gives you orders of magnitude more range.
This allows much more detail to be preserved in the brightest and
darkest parts of the image, where the 8-bit rgb values tend to alias
and clip at the extreme ends, which is lossy.
-Casey
