bulia byak wrote:
... I love math, too. But I'm not obsessed with it. If I feel that some ad-hoc, highly empirical coefficient can improve the subjective behavior of a tool, I add it without hesitation, even if the result will make a mathematician cringe.
As a result, I also like to consider problems that obviously have no simple mathematical solution and therefore escape the attention of more mathematically-minded developers. For example, it is quite obvious that at this time, it is impossible to create a fully automatic filter that would magically turn any halftone image into a decent-looking line engraving; this requires a level of AI beyond what is available these days. But I didn't stop at this realization - and just tried to think up some automatic tools that could make this, still largely manual, work easier.
I prefer looking at this differently, mathematics (as well as computer science) gives you lots of tools to do amazing things. Unfortunately many people seem to make the mistake of seeing mathematics as a solution instead of a tool.
Anyway, I thought you might like to have a look at: http://pages.cpsc.ucalgary.ca/~isenberg/projects/hqrendering.html He uses non-photorealistic rendering techniques to produce high quality vector renderings (apparently to SVG) of 3D shapes, it's not 100% automatic if I remember correctly, but it's probably nearly as close to automatic as you'd want to get (in this area you'll always want to have some room left for tweaking the end result to achieve different effects).