Well, I’ve found my long winters-eve, and made some good progress. I was able to clean up the code to the point that tackling the UV coordinate problem has become … well … possible =) I’ve splitted the algorithm in two parts. The first does the does the triangle splitting (using simple averages for new edge vertices), the second does the actual smoothing using a single mask for all vertices (except for the borders). This seems trivial, but the original algorithm is described as doing subidivision and smoothing at the same time, using different masks for odd and even vertices.

Anyway, I’m still facing some problems, but I promise I’ll post the results when I’m done =)