Remeshing algorithm for multiresolution prior model in segmentation

We consider the construction of a multiresolution triangular mesh for 3D model-based segmentation. In the 3D Snake approach this model needs to be both uniform and regular for convergence purposes. We propose here a remeshing algorithm to compute such a model from an arbitrary triangular mesh. The input mesh is first parameterized in 2D. The output mesh connectivity is fixed by a subdivided base mesh. Its geometry comes from the input mesh via the parametric domain tessellation with respect to uniformity constraints. Experiments were carried out with both synthetic and organ models. They prove the low geometry remeshing error and the uniformity of the resulting mesh.