Exploring feature detection techniques for time-varying volumetric data

The fundamental purpose of scientific visualization is to help scientists extract information from large volumetric datasets. These multi-dimensional datasets may be either derived from observations or generated by simulations. In either case, visualization directly enhances scientific discovery, assists the validation and verification of simulation models, and helps study and predict phenomena. Although the state-of-the-art visualization techniques allow insightful presentations of datasets in various ways, the ability to discern significant features from complex data is lacking. On the other hand, lots of work has been done in the computer vision field, in attempting to automatically detect and recognize features or regions of interest in two-dimensional image data. How to extract features or locate regions of interest in visualizing high-dimensional datasets is an important area of research. We present the work we have done in exploring feature extraction techniques for time-varying three-dimensional volumetric datasets. We used an edge detection method and exploited both temporal and spatial coherences inside features to automatically locate and track the feature movement over time. The results are attractive and show that feature extraction techniques could greatly enhance visualization procedures