School of IT Technical Report CONTOUR TREE SIMPLIFICATION BASED ON A COMBINED APPROACH

The contour tree is a topological abstraction of a scalar field. It represents the nesting relationships of connected components of isosurfaces or contours. The real-world data sets produce unmanageably large contour trees because of noise and artifacts. This makes the contour tree impractical for data analysis and visualization. A meaningful simplification is necessary to the contour tree. This paper proposes an importance-driven approach which combines different measures of importance into a single contour tree simplification pipeline through an importance triangle (ITri). The ITri is set up based on a importance measure vector, whose components are different measures of importance. The Extended Gaussian Image (EGI) and a map projection method are then used to map a importance measure vector to a point in a 2D plane. The points corresponding to importance measure vectors in the 2D plane help the user to differentiate branches of interest from others. Interaction interfaces are provided to manipulate this one-to-one mapping. Through the interaction interfaces, the user can specify models for combining different measures of importance into the contour tree simplification pipeline. The proposed importance-driven and user-directed contour tree simplification approach shows advantages compared with traditional approaches: 1) it combines different measures of importance into a single contour tree simplification pipeline; 2) it makes full use of advantages and suppress disadvantages of each measure of importance; 3) the combined measure of importance is user-directed through different interfaces; 4) the proposed approach allows the user to find patterns of priority values with the contour tree priority curve window, which are used to improve the simplification process.