Stevens Creek Software LLC
Presently it is possible to extract 3D point clouds from complex 3D scenes, often at real-time video rates. What are needed automatic and computationally stable methods to extract geometry from noisy non uniform clouds. These methods must handle automatically under-sampled regions, for example, they have to deal with surface boundaries and resolve the ambiguities imposed by objects that are or appear to be tangent to each other. Furthermore, given the expected low quality of the data it is crucial to have methods to evaluate the consistency of the recovered geometry even in the absence of ground truth.