Visualization Methods for Point Data in Space

"Point-cloud" or "scattered-data" visualization is becoming increasingly important in new emerging applications, especially in sensor network data analysis. Advances in wireless sensor networks are producing more and more data at random points in space and time that must be processed to make possible meaningful three-dimensional visualization, possibly changing with time depending on the specific phenomenon being monitored. Typical variables that can be monitored with sensor networks are temperature, humidity, and light intensity. One goal is to produce visualizations of the monitored variables - but rendered as smoothly varying variables over the particular region of interest. Most scientific visualization techniques require data to include connectivity information, which is not provided by a scattered data set. Techniques used to deal with such unconnected data include the use of field reconstruction methods producing an analytical definition that is then resampled to a standard grid format supported by standard visualization methods, such as volume rendering and iso surfacing. While these methods work well for offline analysis, they are less practical for real-time visualization and become even less effective as data size increases. Highly efficient schemes operating directly on raw scattered data are necessary. We are developing methods for the direct rendering of scattered data, which involves using different building blocks to construct high-quality data visualizations. For instance, the construction of iso surfaces typically is done by extracting isotriangles from some spatial grid structure. Iso surfaces can also be constructed directly by merely using point primitives, given in space with associated function value but no connectivity information. We are developing prototypes for the rendering of scattered data sets using point primitives directly, without performing any meshing steps. Our method is called "iso-splatting," and it is a powerful alternative to traditional extraction-based iso surface visualization. We plan to generalize other visualization techniques as well, so we can use them directly for scattered data, especially massive amounts of sensor data.