We present multiresolution B-spline surfaces of arbitrary order defined ove
r triangular domains. Unlike existing methods, the basic idea of our approa
ch is to construct the triangular basis functions from their tensor-product
relatives in the spirit of box splines by projecting: them onto the baryce
ntric plane. The scheme works for splines of any order where the fundamenta
l building blocks of the surface are hierarchies of triangular B-spline sca
ling functions and wavelets: spanning the complement spaces between levels
of different resolution. Although our basis functions have been deduced fro
m the corresponding 3D bases, our decomposition and reconstruction scheme o
perates directly on the triangular mesh using hexagonal filters. The result
ing basis functions are used to approximate triangular surfaces and possess
many useful properties, such as multiresolution editing, local level of de
tail, continuity control, surface compression, and many more. The performan
ce of our approach is illustrated by various examples, including parametric
and nonparametric surface editing and compression.