A Deformable Modeling Paradigm for 3D Shape Recovery from Visual Inputs

In this paper, we propose a deformable modeling paradigm for 3D shape recovery from visual inputs such as volumetric data and 3D point clouds. The new model is capable of automatically evolving its shape to capture the geometric boundary of the data and simultaneously discover its underlying topological structure. The deformation behavior of the model is governed by partial differential equations (e.g. the weighted minimal surface flow) that are derived by the principle of variational analysis. Unlike the level-set approach, our model always has an explicit representation of geometry and topology. The regularity of the model and the stability of the numerical integration process are ensured by three mesh optimization techniques and the Laplacian tangential smoothing operator. By allowing local adaptive refinement of the mesh, the model can accurately represent fine details. The results of our extensive experiments on synthetic and real data demonstrate robustness, high reconstruction accuracy and visual quality. 1