Awesome! Is there a reference implementation for this? I think it'd be really cool to try to run a complex game like Super Mario or Doom entirely in the terminal using version of this algorithm.
A lot of the Aasci generators seems to try and reproduce the color gradients - would a similar effect to this not happen of we try to reduce the image to line-art first?
Here is a processing program to do ASCII art that has depth info, color, and is animated - http://pastebin.com/bz1pKAKQ
The logic is pretty simple to translate to other languages. This doesn't do the boundary or structural stuff as described in the paper but is a reasonable start for someone interested. Personally I don't like that sparse looking style of ascii art.
Also, I would suggest looking into Tien-Tsin Wong's other works [2] as well. A lot of great ideas and papers there.
[1] https://appsrv.cse.cuhk.edu.hk/~ttwong/papers/asciiart/ascii... [2] https://appsrv.cse.cuhk.edu.hk/~ttwong/publication-favorite....