Better anti-aliasing

Paul Fox pgf at foxharp.boston.ma.us
Fri Apr 25 09:07:14 EDT 2008


bert wrote:
 > 
 > There is no need for "fancy color-adaptive subpixel rendering". The  
 > framebuffer with its 1200x900 resolution maps 1:1 to physical display  
 > pixels. The DCON simply selects the red channel of the first pixel,  
 > and the green of the second, and the blue of the third, and so on.  
 > This does not affect pixel geometry. Thus normal full-pixel anti- 
 > aliased software rendering will do The Right Thing.

i think there's something odd with this description.  (and, as
someone who has clearly been confused about the display for some
time, i'd like to be sure i understand.)

unlike a traditional display, every pixel has a single color. 
given this, it seems wrong to talk about the "red channel of the
first pixel".  you either use some of that pixel, or you don't. 
in effect, the display is implementing sub-(full-color-)pixel
rendering all by itself.  (which i think is what bert was saying,
but the "channel" thing confused me.)

with the backlight off, this doesn't matter -- all pixels are the
same color anyway.  but with the backlight on in monochrome mode,
you see a lot of color in single-pixel-width lines.  if the line
is perfectly aligned with the diagonal of the pixels, it will be
completely colored (i just tried it), but mostly you just get a
lesser or stronger fringing effect.

for backlit ebook mode, this might not matter so much, since ebooks
might not have a lot of diagonal single-pixel-width lines, but the
effect is definitely there.

paul
=---------------------
 paul fox, pgf at foxharp.boston.ma.us (arlington, ma, where it's 58.3 degrees)



More information about the Devel mailing list