Sik, MartinKrivanek, JaroslavBruno Levy and Xin Tong and KangKang Yin2014-01-272014-01-272013978-3-905674-50-7https://doi.org/10.2312/PE.PG.PG2013short.017-022We present a simple and fast algorithm for generating randomly distributed points on a triangle mesh with probability density specified by a two-dimensional texture. Efficiency is achieved by resampling the density texture on an adaptively subdivided version of the input mesh. This allows us to generate the samples up to 40 x faster than the rejection sampling algorithm, the fastest existing alternative. We demonstrate the algorithm in two applications: fast placement of hair roots on a surface and sampling of illumination from a complex luminaire. Part of our mesh sampling procedure is a new general acceleration technique for drawing samples from a 1D discrete probability distribution whose utility extends beyond the mesh sampling problem.I.3.5 [Computer Graphics]Computational Geometry and Object ModelingSurfaces and object representationsI.3.7 [Computer Graphics]Three Dimensional Graphics and RealismColorshadingshadowingand texture