Intrinsic Point Cloud Simplification

Modelling and visualisation methods working directly with point-sampled geometry have developed into attractive alternatives to more traditional mesh-based surface processing. In this paper, we consider a vital step in any point-based surface processing pipeline, point cloud simplification. Building upon the intrinsic point cloud simplification idea put forward in [14], we obtain a simplification algorithm allowing for intuitive density control and satisfying a set of important requirements unsupported by existing simplification techniques. The algorithm operates efficiently and gives a point set density guarantee. It supports both sub- and resampling of the input point set and allows for uniform and user-controlled feature-sensitive simplification. It can further deal with non-uniformly distributed point sets and point-sampled geometry featuring illegitimate holes of simple complexity. The algorithm is inherently progressive and supports the generation of multiresolution representations in the form of levels of detail. We are primarily concerned with describing the conceptual framework of our intrinsic approach and show its viability by giving a number of application examples using massive data sets