Void-and-Cluster Sampling of Large Scattered Data and Trajectories.

Area of Science:

• Computer Science
• Data Science
• Scientific Visualization

Background:

• Scattered data presents challenges for efficient storage and visualization.
• Existing data reduction techniques may not adequately capture spatial distribution or adapt to data complexity.

Purpose of the Study:

• To introduce a novel statistical sampling technique for scattered data reduction.
• To enable adaptive sampling based on data density and ensure optimal spatial distribution.
• To support progressive data loading and continuous level-of-detail representations.

Main Methods:

• Void-and-cluster sampling technique for optimal spatial distribution (blue noise property).
• Adaptive sampling to density functions for prioritizing complex data regions.
• Extension to time-dependent trajectories (e.g., pathlines) using an iterative approach.
• Development of a local and continuous error measure for representation quality.

Main Results:

• The technique generates a representative subset with blue noise properties.
• Adaptive sampling allows denser sampling in high-complexity multivariate value domains.
• Implicit sample ordering facilitates progressive loading and level-of-detail.
• The error measure effectively guides sampling and quantifies representation accuracy.

Conclusions:

• The proposed void-and-cluster sampling is an effective data reduction technique for scattered and time-dependent data.
• The method offers advantages in spatial distribution, adaptive complexity handling, and progressive visualization.
• The introduced error measure provides valuable insights into sampling quality, performance, and scalability.