On Thu, 2009-03-26 at 11:51 +0100, Roberto Ragusa wrote:
Just a thought: do I remember correctly that the color filter is just
a
simple y(0)=(x(-1)+x(0)+x(1))/3 on the subpixels? Did anyone consider that
R, G and B have different brightness to the human eye and should be weighted
more smartly?
Everyone's favorite self-aggrandizing lunatic Steve Gibson has this
algorithm:
http://www.grc.com/cttech.htm
Which looks quite good to me and makes a lot of sense. But playing with
his little applet I really don't see the advantage of the second pass,
the first filter pass looks just fine to me, I can hardly see the
difference the second pass makes, it just looks very slightly blurrier.
Note that his applet only seems to be working with un-antialiased text
with good hinting, which probably makes a difference. If your underlying
font rendering looks like crap, fringe filtering isn't going to help it
any. Garbage in garbage out.
I have no idea how Gibson's algorithm relates to any patents. Gibson
doesn't seem to have any, but a lot of patents seem to reference his
page...