Interactive Structure-aware Blending of Diverse Edge Bundling Visualizations

Yunhai Wang*1    Mingliang Xue1    Yanyan Wang1    Xinyuan Yan1   
Baoquan Chen2    Chi-Wing Fu3    Christophe Hurter4
1Shandong University   2Peking University  3The Chinese University of Hong Kong 
4French Civil Aviation University  

Accepted by IEEE InfoVis 2019


Figure 1: Transplanting the sub-layout selected by the black box in (a) to the corresponding region of the unbundled layout in (b). (c) Directly replacing the selected regions results in jagged boundaries; (d) smoothing the layout in (c) results in some bundled structures missed; (e) result produced by our method using the constraints of structural preservation and smoothness; and (f) result produced by our method further using the readability constraints that relieve visual ambiguity, where the upper and bottom boxes highlight the result changes before and after applying this constraint.



Abstract:

Many edge bundling techniques (i.e., data simplification as a support for data visualization and decision making) exist but they are not directly applicable to any kind of dataset and their parameters are often too abstract and dificult to set up. As a result, this hinders the user ability to create eficient aggregated visualizations. To address these issues, we investigated a novel way to handle visual aggregation with a task-driven and user-centered approach. Given a graph, our approach produces a decluttered view as follows:first, the user investigates different edge bundling results, and specfies areas, where specfic edge bundling techniques provide good results. Second, our system then computes a smooth and structural preserving transition between these specfied areas. Lastly, the user can furtherfine-tune the global visualization with a direct manipulation technique to remove the local ambiguity and to apply different visual deformations. In this paper, we provide details for our design rationale and implementation. Also, we show how our algorithm gives more suitable results compared to current edge bundling techniques, and in the end, we provide concrete instances of usages, where the algorithm combines various edge bundling results to support diverse data exploration and visualizations.


Video:




Materials:

Paper: [PDF].


Results:

     

Figure 2: These visualizations are created using different edge bundling methods and rendering styles for presenting recorded aircraft trajectories over France during one day. From Left to right, the visual simplfication varies from no simpli?cation to a signficant one. The width of edge segments encodes the edge density, and the top row displays the edge segments with high densities on top so that the main bundles are more clearly shown, while the bottom row displays the edges in a reverse way. The last column shows a strong aggregation with high transparency, where the main air flows are more clearly shown. From these various rendering and edge bundling methods, the user can decide which part of each visualization best shows the relevant information and combine them together by using our blending method. For example, the regions with labels 𙹗 are selected to blend with the original layout with the label 1 together, formed the result shown in Figure 3(b).

     

Figure 3: (a) Original data of one-day aircraft trajectory record over France. Red boxes show areas, where our method blended different edge bundling results from Figure 2 to compose the final image shown in (b). (b) Final result produced by our blending method, which takes into account the selected area (red boxes in (a)) using different renderings and edge bundling parameters. Our method insures smooth transitions between areas and also avoids ambiguity.

     

Figure 4: This example shows the usage of the origin-destination feature. (a) The original US migration data-set with the links in yellow between the selected source and destination nodes, and the bounding box of these links in blue; (b) the bundled result generated by applying the KDEEB algorithm [17] to (a); and (c) further aggregated result by our method, where the ambiguity between edges shown in the black box is resolved.

     

Figure 5: (a) Input unbundled graph. (b) Cluster fisheye lens zoom-in on the red cluster with black border by (c) bundling the context area while keeping the focus area unbundled. Route the unrelated edges, which go through the focused cluster. (d) Multi-focus fisheye lens on the red and purple clusters, and bundle the associated context area.


Acknowledgment:

This work is supported by the grants of the National Key Research & Development Plan of China (2016YFB1001404), NSFC (61772315, 61861136012) and the Research Grants Council of the Hong Kong Special Administrative Region (Project no. CUHK 14203416) and the French National Agency for Research (Agence Nationale de la Recherche ANR) under the grant ANR-14-CE24-0006-01 project 揟ERANOVA.