Towards a multi-level modeling approach for reconstruction. Application to medical data

This paper introduces a new modeling approach we call multi-level, which is well adapted to smooth objects coming from medical data. We believe in the importance of taking into account the global and the local characterization of an object, within a single model. Our model takes advantage of implicit surfaces with skeletons to represent the shape globally. Successive implicit layers then refine the model, in order to describe each level of detail in a coherent way, until the finest level, which corresponds to small local variations of the shape. We also present an original reconstruction method based on this model. Finally an application to the vertebra is presented.