2D Tensor Field Segmentation

Authors Cornelia Auer, Jaya Sreevalsan-Nair, Valentin Zobel, Ingrid Hotz

Thumbnail PDF


  • Filesize: 3.25 MB
  • 19 pages

Document Identifiers

Author Details

Cornelia Auer
Jaya Sreevalsan-Nair
Valentin Zobel
Ingrid Hotz

Cite AsGet BibTex

Cornelia Auer, Jaya Sreevalsan-Nair, Valentin Zobel, and Ingrid Hotz. 2D Tensor Field Segmentation. In Scientific Visualization: Interactions, Features, Metaphors. Dagstuhl Follow-Ups, Volume 2, pp. 17-35, Schloss Dagstuhl – Leibniz-Zentrum für Informatik (2011)


We present a topology-based segmentation as means for visualizing 2D symmetric tensor fields. The segmentation uses directional as well as eigenvalue characteristics of the underlying field to delineate cells of similar (or dissimilar) behavior in the tensor field. A special feature of the resulting cells is that their shape expresses the tensor behavior inside the cells and thus also can be considered as a kind of glyph representation. This allows a qualitative comprehension of important structures of the field. The resulting higher-level abstraction of the field provides valuable analysis. The extraction of the integral topological skeleton using both major and minor eigenvector fields serves as a structural pre-segmentation and renders all directional structures in the field. The resulting curvilinear cells are bounded by tensorlines and already delineate regions of equivalent eigenvector behavior. This pre-segmentation is further adaptively refined to achieve a segmentation reflecting regions of similar eigenvalue and eigenvector characteristics. Cell refinement involves both subdivision and merging of cells achieving a predetermined resolution, accuracy and uniformity of the segmentation. The buildingblocks of the approach can be intuitively customized to meet the demands or different applications. Application to tensor fields from numerical stress simulations demonstrates the effectiveness of our method.
  • Tensorfield visualization
  • surface topology


  • Access Statistics
  • Total Accesses (updated on a weekly basis)
    PDF Downloads
Questions / Remarks / Feedback

Feedback for Dagstuhl Publishing

Thanks for your feedback!

Feedback submitted

Could not send message

Please try again later or send an E-mail