An Approach for 3d Segmentation on Multiresolution Surfaces

We present an approach for 3D segmentation on multiresolution surfaces of 3D data that are represented in the form of triangular meshes. In this approach, multiresolution surfaces are generated by applying two algorithms, mesh decimation and surface fairing methods. The decimation method simplifies surfaces into less complex forms which contain less details than the original surface. The surface fairing method is for smoothening the simplified surfaces. Patches of different multiresolution surfaces are then generated based on geometry and differential geometry information. Our approach is useful for the problem of surface matching, an important problem in computer vision. In this paper, we show experiments on two kinds of data, 3D face data and 3D protein structure data.