The Method of Mixed States for Interactive Editing of Big Point Clouds

Abstract

We present a novel methodological approach for the interactive editing of big point clouds. Based on the mathematics of fiber bundles, the proposed approach to model a data structure that is efficient for visualization, modification and I/O including an unlimited multi-level set of editing states useful for expressing and maintaining multiple undo histories. Backed by HDF5 as high performance file format, this data structure naturally allows persistent storage for the history of modification actions, an unique new feature of our approach. The challenges of visually based manual editing of big point clouds are discussed and a proper rendering solution is presented. The implemented solution and its features as consequences of the underlying methodology is compared with two major mainstream applications providing point-cloud editing tools as well.