Invariant crease lines for topological and structural analysis of tensor fields

Area of Science:

• Scientific visualization
• Computer vision
• Medical imaging
• Engineering research

Background:

• Characterizing complex 3D symmetric second-order tensor fields is challenging.
• Topological approaches define degenerate lines, but their extraction can be difficult.
• Identifying salient structures is crucial in fields like diffusion tensor MRI and computational mechanics.

Purpose of the Study:

• To introduce a versatile framework for characterizing and extracting salient structures in 3D symmetric second-order tensor fields.
• To reformulate the extraction of topological lines in tensor fields using tensor invariants.
• To enable the application of existing computer vision and scientific visualization techniques to tensor field analysis.

Main Methods:

• The framework reformulates degenerate lines in tensor fields as crease lines of the tensor invariant 'mode'.
• It utilizes well-established methods from scientific visualization and computer vision for line extraction.
• Implementation involves smooth reconstruction kernels and adaptive refinement for non-linear scalar measures and smooth manifold computation.

Main Results:

• Demonstrated that degenerate lines in tensor fields correspond to crease lines of the 'mode' invariant.
• Showcased the framework's ability to extract important structural properties using other tensor invariants like fractional anisotropy (FA).
• Successfully applied the method to diffusion tensor MRI data and benchmark engineering stress tensor fields.

Conclusions:

• The proposed framework offers a robust and versatile approach for analyzing 3D tensor fields.
• It bridges topological methods with tensor invariant analysis, enabling new applications.
• The method facilitates the identification of critical structures in complex scientific and engineering datasets.