This looks impressive, but does anybody know why "exact coverage" is apparently considered the gold standard for rendering vector graphics? Mathematically, computing pixel coverage corresponds to sampling a box filtered version of a characteristic function.
In practice I would expect, say, a gaussian filter to be both easier to approximate and less prone to aliasing artifacts. Apparently that expectation is completely wrong though, since nobody seems to implement it that way! What's so special about vector graphics that makes the box filter behave well?
In practice I would expect, say, a gaussian filter to be both easier to approximate and less prone to aliasing artifacts. Apparently that expectation is completely wrong though, since nobody seems to implement it that way! What's so special about vector graphics that makes the box filter behave well?