Re: 24-bit-per-pixel limit to Gdk RGB ?

[Peter Finderup Lund <firefly diku dk>]

On Tue, 18 Feb 2003, individual wrote:

> Ok, last attempt: I made a 1000 pixel wide gradient going from 255 red
> to 254 red. In the colour selection panel, I can clearly see the
> difference between the two colours. When I draw the gradient, there is
> no dividing line. HA! I have therefore proved that The Gimp and my X
> server  are capable of displaying colour at more than 8 bits per
> channel. Or have I?

Did you do this experiment with the gimp?  In that case you proved that
the gimp does use dithering to avoid Mach banding.

Almost no X servers on Earth will currently give you more than 8 bits per
colour channel.  I don't know which X server you are using and what
graphics card you are using.  A few very new consumer cards will give you
10-bits per colour channel (or is it only 10-bits for the green channel?)
and I think XFree86 can be hacked/is being hacked to support this.  It
does not, however, have the infrastructure in place for more than 32 bits
for the colour channels + alpha combined.  If you need that, you will have
to get special equipment, probably from Sun or SGI (or Evans & Sutherland,
if they still exist -- btw. look up their names in citeseer).

There is no point in putting support for dithering 48-bit to 24-bit pixels
into GdkRgb since it is only needed in very specialized situations.  And
if you really need it that much you can probably afford to implement it
yourself.

So, do you have a normal photographic image in 48-bit colour that you want
to display or do you have some specialized purpose?

In the first case, you can throw away (more than) half the bits with out
anyone being the wiser.  In the second, you will have to be creative.

google for something like error diffusion dithering + some more of Raph
Levien's work.

-Peter

"Of course, I'm not unbiased, but in my humble opinion, I've
 gotten close to something that I can be really proud of."
 -- Knuth on The Art of Computer Programming.