I have just committed a new LCD rendering technique, which has been in
the Harmony branch for some time now.
Basically, forget tripling the resolution, forget LCD low-pass
filtering. Instead render 3 grayscale bitmaps by shifting the outline
by 1/3 pixel and combine the results. You are done without any
additional filtering. The result is supposed to be identical to the
light filtering using the ClearType technique
I have been soliciting feedback for a few month now specifically
regarding the Clear Type patents. Some folks pointed out this specific
paragraph: "The color values are independently generated for each of
the red, green, and blue pixel sub-components based on different
portions of the image, rather than the color values for the entire
pixel being generated based on a single sample or the same portion of
the image." This is indeed fundamental to ClearType, where each color
channel gets coverage information from completely *different*
(distinct, non-overlapping) portions of the glyph. It is too bad that
it also results in *unequal* coverage of each channel grid (aka color
fringes) and filtering becomes necessary to equalize the channels.
So, Harmony is equalized to begin with because shifting does not
change the integrated channel coverage. As for the patents, the
adjacent channels get coverage from parts of glyph that have 2/3
overlap (truly, more similar than different). This is how I would
argue that this technique is substantially not ClearType.
What do you think?
Freetype-devel mailing list