So now let's walk through the algorithm in action. First let's see how a
high-resolution texture would be stored in a silhouette map texture.
Here we have some vector artwork which shows a green polygon
in front of a yellow background.
If we simply converted this into a standard texture using the quantization
indicated we would get something
that would not really capture the sharp edges of the polygon: we can either
perform nearest-neighbor sampling of this original artwork to get a
pixelated texture or we can
perform filtering when we downsample to get a fuzzy (but antialiased) texture.
Neither one of them would magnify properly, however.
Silhouette maps, on the other hand, will correctly capture the necessary
detail. Here, the red dots
indicate silhouette points on
the discontinuity and the orange-pink dots are the default points that have been
placed in the middle of each cell. The orange squares show the cells of the
silhouette map, and the squares on the corners of the cells indicate the RGB
color samples.