Donald Bren School of Information and Computer Sciences

With the fast increasing size of captured 3D models, i.e. from high-resolution laser range scanning devices, it has become more and more important to provide basic point processing methods for large raw point data sets. In this paper we present a novel stream-based point processing framework that orders unorganized raw points along a spatial dimension and processes them sequentially. The major advantage of our novel concept is its extremely low main memory usage and its applicability to process very large data sets out-of-core in a sequential order. Furthermore, the framework supports local operators and is extensible to concatenate multiple operators successively.