A Fast and Robust Method for Visualizing Separation Line Features Xavier Tricoche 1 , Christoph Garth 2 , and Gerik Scheuermann 3 1 Scientific Computing and Imaging Institute, University of Utah tricoche@sci.utah.edu 2 Department of Computer Science, University of Kaiserslautern garth@rhrk.uni-kl.de 3 Institute of Computer Science, University of Leipzig scheuermann@informatik.uni-leipzig.de The visualization of a three-dimensional viscous flow around an embedded object is typically based on the analysis of its wall shear stress. This vector field defined over the object body exhibits structures that are key to the qualitative evaluation of the surrounding flow. Open separation and attachment lines are of essential interest in aerodynamics due to their adverse effects on the object motion and their impli- cation in vortex genesis. The paper presents a new method for the efficient analysis and visualization of separation and attachment lines on polyhedral surfaces in three- space. It combines local prediction and global feature extraction to yield a scheme that is both efficient and accurate. In particular, it does not suffer from the restrictions induced by assumptions of local linearity and is able to detect features where exist- ing techniques fail. The algorithm is built upon an efficient streamline integration scheme on polyhedral surfaces. The latter is also employed to develop a variation of the LIC scheme. Results are proposed on CFD data sets that demonstrate the ability of the new technique to precisely identify and depict interesting structures in practi- cal applications. 1 Introduction Modern numerical simulations in Computational Fluid Dynamics (CFD) generate large scale datasets that must undergo qualitative and quantitative evaluation for in- terpretation. Typically, the analysis relies on the extraction and identification of struc- tures of interest that are used to gain insight into essential properties of the flow for the considered application. In the field of aircraft design in particular, huge amounts of flow data are computed and processed to better understand the properties of de- sign prototypes or to look for optimal configurations, especially during critical flight situations. The usual approach to this analysis is to study the interaction between the three-dimensional air flow around the body and the so-called shear stress vector