Preliminary release of Squeak3.0 available for test

Ted Wright wright at en.com
Tue Feb 6 00:23:59 UTC 2001


Lawson English <english at primenet.com> wrote:
>on 2/5/01 10:21 AM, Steve Wessels at frazier at one.net wrote:
> > True color is 24 bits.
> > 
> > I think this comes from the idea that the eye cannot resolve more than
> > 24 bits of color.
>
>
>Which isn't true. There are natural images where banding shows up with only
>24-bits of RGB, and it is trivial to contrive a banded 24-bit image.

I think you are both right. As I recall, the human eye can only distinguish about
200 distinct shades of grey, and is less sensitive to color (which is two more
channels in addition to grey). So it seems like three 8-bit (256 value) channels
should be enough.

The catch is that the 200 grey levels we can distinguish are not linearly spaced
they way binary numbers are. They need to be logarithmically spaced if you only
want to use 200 values. You need 12 bits of linearly spaced values (per channel) 
to represent the full range of human vision. 

I've heard this is why some digital camera are 12 bits per plane internally even 
though they output only eight. They play games with gamma adjustments or 
something like that to increase the range covered by the camera (at least that's
what the guy from Kodak told me).

Ted


Ted Wright
NASA business: mailto:wright at grc.nasa.gov    remainder: mailto:wright at en.com





More information about the Squeak-dev mailing list