This article is the third and final in a series of wrap-ups from VizWorld reader T.J. Jankun-Kelly (@dr_tj) of the EuroVis2010 conference currently underway.  If you missed it, see Part 1 and Part 2.

The last day of the conference saw an abbreviated program with two sessions and the plenary. The Best Paper was also awarded. As with the previous day, there were two tracks.

Session 7B: Vectors & Tensors

An Evaluation of Glyph Perception for Real Symmetric Traceless Tensor Properties; T.J. Jankun- Kelly, Yagneshwara Lanka, Edward Swan

This presentation by your humble scribe discussed a perceptual study involving tensors from liquid crystal physics. The motivation was to determine if the glyphs we had previously designed communicated the three perceptual dimensions of the data (orientation, compression/elongation, and flatness) better than other glyphs used by physics. Two users studies involving 30 subjects found that our superquadric tensor glyphs were significantly more accurate in communicating these features; cylindrical glyphs faired the second best with ellipsoids and box-like glyphs following. Similar studies could be performed for other tensor glyphs to determine their efficacy.

Read the read of TJ’s writeup after the break.

A Maximum Enhancing Higher-Order Tensor Glyph; Thomas Schultz, Gordon Kindlmann

Shutlz presented work done with Kindlmann at Chicago regarding higher-order tensor glyphs. They have designed a glyph with properties similar to ellipsoids in the higher-order space in that it clearly identifies orientations of multiple axes in such tensors. Color is used to indicate orientation in addition.

Topology Aware Stream Surfaces; Dominic Schneider, Wieland Reich, Alexander Wiebel, Gerik Scheuermann

Streamsurfaces are a continuous means to depict the shape of a tensor or vector field; however, they suffer near saddle critical points in that they tend to “tear” or separate due to the high rate of change near such extrema. Reich proposed a method that detects a linear vector field approximation around such extrema and adapts the streamsurface generation in this region to eliminate the visual discrepancies.

Toward a Lagrangian Vector Field Topology; Raphael Fuchs, Ronny Peikert, Jan Kemmler, Benjamin Schindler, Juergen Waser, Filip Sadlo, Helwig Hauser

Unsteady flow is complex to visualize, especially due to the fact that highly unsteady flow cannot be modeled using vector field topology beyond fields with Galilean-like behavior. Piekert discussed several metrics to extract Lagrangian topology that are more robust in highly unsteady flows. A combination of two metrics proved best for their tests, and further refinements are being pursued.

Session 8A: Evaluation

The Readability of Path-Preserving Clusterings of Graphs; Daniel Archambault, Helen C. Purchase, Bruno Pinaud

Archambault discussed a thorough study examining the difference between non-clustered and clustered node-link diagrams where the clusters are represented as meta-nodes (nodes of nodes). Connectivity, classification, and two different path-related tasks were used over 20 users. Results were found to be largely size dependent, with significant differences in mid-sized graphs as opposed to small or large ones; other similar results were reported for the other tasks.

A Salience-based Quality Metric for Visualization / Heike Jänicke, Min Chen

Jänicke & Chen from Swansea explore the use of visual salience as a metric to evaluate visualizations. Assuming an a priori importance measure has been given by an expert, they compare this visualization saliency versus an image saliency measure calculated by computer vision algorithms. They can then classify areas of high, medium, or low matching between the two saliences and indicate where too much visual input or given or where too little is provided. The image saliency can also be used by itself to provide a post hoc indication of visual importance that could be used to improve visualization design.

Understanding Interactive Legends: a Comparative Evaluation with Standard Widgets; Nathalie Henry Riche, Bongshin Lee, Catherine Plaisant

Riche discussed work with Lee & Plaisant regarding the use of interactive legends to facilitate visualization. While legends are needed to provide readable visualizations, they explore the design constraints needed for legends that are interactive. They classify four different kinds of legends based upon the type of data (nominal, ordinal, and so on) and evaluated the effectiveness of the legend in comparison to standard dynamic query siders.

The Perception of Correlation in Scatterplots; Ronald Rensink, Gideon Baldridge

The final talk of the section and this part of the conference, Resnik outlined how long standing rigorous methods form vision science can be used to evaluate information visualization. Resnik outlined an experiment of correlation in scatterplots, discussing how standard visual science methods were used. A Weber-like law of visual perception of correlation was discovered and further guidelines for such experiments were provided.

Closing Session

EuroVis concluded with awards and details on EuroVis 2011. The next iteration of EuroVis will be in Norway in Bergen, hosted by Helwig Hauser and his group at UiB. It will be held May 31–June 3rd, with EuroVAST to precede it.

For the best papers, three papers were chosen in rank order. Topological Landscapes by Harvey & Wang was selected as 3rd Best due to its novel characterization of the topology of high dimensional spaces. Following was Image-based Edge Bundling work of Telea and Orsoy due to its novel improvements of a well used algorithm. Best Paper, however, was awarded to Bruckner & Möller for Isosurface Similarity Maps; the committee felt that this characterization of isosurface shape space solved an important problem in a clear manner with an opaque discussion of its limitations. Congratulations to all the winners!