The rendering technique is based off of a specific tile design. A little ASCII art is necessary:
oooooooo
oooooooo
oooooooo
oooooooo
oooooooo
oooooooo
oooooooo
oooooooo
This is an 8x8 pixel tile. There are two color values, and they may be alternated. There are 3 more patterns that are rotations of this pattern, there are 2 patterns with the two smaller squares in opposing corners, and 1 blank tile with no small squares. With all small-square orientations and color alternations, there are a total of 14 tiles.
Without the small squares, there is no illusion, the illusion springs from the small squares in a way that I only understand intuitively; I cannot quite articulate it. All I can really say is that it sort of tricks the mind into seeing an aliased line, the stair-stepping that happens when rendering a solid line on a raster display without an anti-aliasing algorithm.
With the 8x8 tile, the small squares MUST be this size, any larger and they are no longer visible as squares (they would appear as half of the tile); any smaller and they are single pixels and too small. However, with larger tiles, like 32x32, the size of the smaller squares can be modulated. This may enable us to indicate areas of differing slope. The direction of the slope selects the tile pattern and the magnitude of the slope selects the size of the small squares.
We'll see.
edit: oh, I got an A on my paper.
I publish my first paper later this summer as well.