Perhaps every odd scanline should concentrate on matching blue and green
values against everything, and every even scanline should concentrate on
matching red values against everything.
In the easiest implementation, what you'd get is something similar to
Colorview for the Atari 8-bit where each scanline is assigned a part of
the spectrum (in its case, red green and blue). But it doesn't have to
be that way.
For the blue-green scanlines, for example... Perhaps choose the
colors based on which popular colors are most blue, and which are
least blue. That could provide forced contrasts between red and
purple, yellow and white, black and cyan, etc. Then you'd be able
to pick up all the differences in blue/green levels in that scanline
without making a pure blue/green scanline. In addition, if the
red scanline doesn't pick apart yellow and white, you'll have some
contrast between it on the next and previous scanlines. The
blue-green scanlines wouldn't be able to differentiate white and
blue, or even black and yellow, but the red scanlines would.
It might be prudent to rotate selection filters over three scanlines
rather than two, because that would yield more accuracy, though it
would probably result in grainier dithering. Also, if we were to
interlace the pictures so that each individual scanline would go
through all the filters, a two-filter system would use less memory
(two screens) and flicker less (30hz vs 20hz).
But realistically, I don't see how full-color pictures can look any
good on the 2600 by using a niversal process. Hopefully my ideas
will help bring us a little closer though.
--
Archives (includes files) at http://www.biglist.com/lists/stella/archives/
Unsub & more at http://www.biglist.com/lists/stella/stella.html
