News Archives

RSS Feed

Multi-variate Volume Visualization

May 2, 2006

  • Date: Tuesday, May 2, 2006 
  • Time: 11:00 am — 12:15 pm 
  • Place: Woodward 149

Joe Kniss  
Computer Science Department University of Utah

Scientific visualization is a discipline that joins data analysis and human visual perception. Visualization addresses a need for high performance, interactive data exploration of ever increasing size and complexity. My work focuses on volume visualization techniques. Volume visualization deals with the discovery and display of important features embedded in three-dimensional data. Volume visualization users are increasingly in need of techniques for assessing quantitative uncertainty and error in the images produced. Statistical segmentation algorithms can compute these quantitative results, yet volume rendering tools typically produce only qualitative imagery via transfer function-based (look-up table) classification. My recent work introduces a visualization technique that allows users to interactively explore uncertainty, risk, and probabilistic decision boundaries. This approach makes it possible to directly visualize the combined “fuzzy” classification results from multiple segmentations by encoding this data in a unified probabilistic data space. This unified space, derived from the combination of scalar volumes from numerous segmentations, is represented using a new graph-based dimensionality reduction scheme. The scheme both dramatically reduces the dataset size and is suitable for efficient, high quality, quantitative visualization. More importantly, this scheme emphasizes the relationship between multi-modal or multivariate image data and the intrinsic nature of “features” encoded in the image. This knowledge can be leveraged to produce high quality visualizations, robust feature detection algorithms, and improved representations of multi-dimensional image data.

This talk will cover a basic introduction to the volume visualization process, graphical methods for rendering and lighting, algorithmic considerations with respect to efficiency and computational hardware, as well as new theoretical results based on this research.