When I started this series, I said that I wanted to find out what kinds of transformation can be applied to line drawings, and what their properties are. I've mentioned one, the texture-deleting one that I call "appointing a representative". Another, the topic of this posting, enhances crucial regions relative to the rest of the drawing. By "crucial regions", I mean those regions that are most important for recognising what the drawing depicts. Every cartoonist who has had to draw the same character at wildly different sizes will, I think, be familiar with this. But for an excellent example and explanation, please read my previous post and Hofstadter's explanation therein of how he renders Helvetica on smaller and smaller grids. The key point there is:
The second and third columns are the work of an algorithm that has information about zones likely to be characteristic and critical for recognizability. It mathematically transforms the original outline so that the critical zones are disproportionately enlarged (the way your nose is enlarged when you look at yourself in a spoon). It then applies the naive algorithm to this new outline (pixels light up if and only if they fall inside). This amounts to an interesting trade-off: sensitivity in the critical zones is enhanced al the sacrifice of sensitivity in less critical zones.