Creative Commons License
The original figures from the book are released here under the Creative Commons Attribution 4.0 International License (CC BY 4.0). When you use a figure for your own work, please, cite the book appropriately, for example, like this: Christian Tominski and Heidrun Schumann. "Interactive Visual Data Analysis". AK Peters Visualization Series, CRC Press, 2020.
Some figures were created by colleagues from the visualization community and are used in the book under the CC BY 4.0 license. Below, these figures are clearly marked by giving the original figure author in the figure caption. When you use these figures, please, do include an appropriate attribution to the original author.
Here is an archive with all figures from the book. Below, you can search for figures based on their caption or browse, view, and download individual figures.
Node-link diagram visualizing a graph's structure and attributes. (a) Plain structure.
Node-link diagram visualizing a graph's structure and attributes. (b) Encoding degree via color.
Node-link diagram visualizing a graph's structure and attributes. (c) Encoding degree via color and size.
Node-link diagram visualizing a graph's structure and attributes. (d) Encoding weight via line width.
Dynamic filtering to focus on relevant parts of a climate network. (a) Full graph with 6,816 nodes and 116,470 edges.
Dynamic filtering to focus on relevant parts of a climate network. (b) Filtered graph with 938 nodes and 5,324 edges.
Multiple-views visualization of a climate network.
Guidance provides assistance during data navigation. (a) Where should I go next?
Guidance provides assistance during data navigation. (b) Visual cues hint at candidates!
Chapter structure of this book.
Visualization of life satisfaction in Germany. (a) Failing visual representation.
Visualization of life satisfaction in Germany. (b) Succeeding visual representation.
Functional dependency between the reference space and the attribute space. For a point in the reference space, there is exactly one point in the attribute space.
Key terms for characterizing data.
The scope defines to which extent an observation is valid. (a) Global scope. (b) Local scope. (c) Point scope.
Visualizing the local scope of measurements of water quality. (a) Data points only.
Visualizing the local scope of measurements of water quality. (b) Voronoi partitioning.
Visualizing the local scope of measurements of water quality. (c) Shepard interpolation.
Meta-data to characterize the data to be analyzed.
Four-set Venn diagram illustrating different classes of data.
Different visual encodings to support different tasks. (a) Coloring suited to identifying values.
Different visual encodings to support different tasks. (b) Coloring suited to locating extrema.
A target as defined by projection and selection.
Goals, questions, targets, and means characterize analysis tasks.
Meteorological measurements over the course of the year. (a) Hours of sunshine.
Meteorological measurements over the course of the year. (b) Air temperature.
Meteorological measurements over the course of the year. (c) Cloud cover.
Histograms of the distribution of cloud cover values. (a) Cloud cover value frequencies in Rostock.
Histograms of the distribution of cloud cover values. (b) Cloud cover value frequencies in Dresden.
Four nested levels outline how to design interactive visual data analysis solutions.
Data-oriented and graphics-oriented stages and operators.
A network of operators describes the data's transformation through several stages from data values to image data.
Knowledge generation model.
Illustration of the effect of different visual representations. (a) Line plot.
Illustration of the effect of different visual representations. (b) Spiral plot (cycle length 32 days).
Illustration of the effect of different visual representations. (c) Spiral plot (cycle length 28 days).
Effectiveness ranking of visual variables.
Color maps for identifying and locating values and classes.
Applying the color maps from Figure 3.4 to temperature data. (a) Color coding for identification tasks.
Applying the color maps from Figure 3.4 to temperature data. (b) Color coding for location tasks.
Basic mapping of a data variable onto a visual variable.
Enhanced data-dependent visual mapping. (a) Value range expansion. (b) Logarithmic mapping. (c) Box-Whisker mapping.
Combined color map for comparing two data variables.
Two-tone coloring explained.
Two-tone visualization of 20 years of daily temperatures.
Visual encoding via position, area, color, and shape. (a) Mapping to position.
Visual encoding via position, area, color, and shape. (b) Mapping to area.
Visual encoding via position, area, color, and shape. (c) Mapping to color.
Visual encoding via position, area, color, and shape. (d) Mapping to shape.
Terrain visualization with overview+detail.
Illustration of focus+context for a table-based visualization of the Iris flower dataset. Focused rows are magnified to accommodate labels. (a) Regular visualization.
Illustration of focus+context for a table-based visualization of the Iris flower dataset. Focused rows are magnified to accommodate labels. (b) Focus+context distortion of rows.
Multiple coordinated views for analyzing multivariate data.
Two-tone colored table-based visualization of the Cars dataset.
Table Lens with textual labels for focused data tuples.
9 × 9 scatter plot matrix of meteorological data. Color is used to ease the recognition of data variables.
Visualization with polylines across parallel and star-shaped axes. (a) Parallel coordinates plot. (b) Radar chart.
Visual patterns between pairs of parallel axes.
The same data as in Figure 3.19 visualized as scatter plots.
Parallel coordinates with histograms showing demographic data.
Examples of classic glyphs for visualization. (a) Autoglyph. (b) Stick figures. (c) Chernoff faces.
Corn glyph for representing six ordinal data values.
Pixel-based visualization of daily values of six meteorological attributes collected for more than hundred years in the city of Potsdam.
Mosaic plot visualizing the survival of passengers of the Titanic.
Temporal relations for time instants and time intervals. (a) Instant relations. (b) Interval relations.
Aspects of time to be considered when visualizing temporal data.
Types of data with references to time.
Small multiples visualization of the number of people diagnosed with problems of the upper respiratory tract.
Time Wheel visualization of human health data.
Visual representation of intervals using the triangular model. (a) Standard interval representation. (b) Intervals in the triangular model.
Stream graph with randomly generated data.
Spiral display with four years of daily temperatures in Rostock.
Comparison of a regular line plot (top) and a cycle plot (bottom).
Visualization of uncertain time intervals for planning purposes.
The TimeViz Browser provides an illustrated overview of more than a hundred techniques for visualizing time and temporal data.
Spatial regions at different scales: Federal state, districts, zip-code regions.
Geo-spatial data can refer to different spatial units. (a) Points. (b) Lines. (c) Areas. (d) Volumes.
Different map projections preserve different spatial properties. (a) Equirectangular.
Different map projections preserve different spatial properties. (b) Mercator.
Different map projections preserve different spatial properties. (c) Natural Earth.
Myriahedral projections of the Earth.
Map representation at different resolutions. (a) Original data 100%.
Map representation at different resolutions. (b) Data reduced to 50%.
Map representation at different resolutions. (c) Data reduced to 10%.
Terrain rendering of the Puget Sound region.
Reducing overlap of stream graph glyphs on a map. (a) Straightforward placement.
Reducing overlap of stream graph glyphs on a map. (b) Overlap-optimized placement.
Indirect visualization of geo-spatial data. (a) Univariate choropleth map plus multivariate parallel coordinates plot.
Indirect visualization of geo-spatial data. (b) Flexible visualization via probes.
Systematic view of 2D and 3D representations of geo-spatial data and geographic space.
3D visualization of the trajectory of an aircraft approaching Sion airport.
Visibility widgets help users identify obscured information in 3D geo-visualizations.
Visualization of movement trajectories. (a) 2D map with 2D paths. (b) 2D map with stacked 3D bands.
Visualizing spatio-temporal data using 3D glyphs on a 2D map. (a) Pencil glyphs for linear trends.
Visualizing spatio-temporal data using 3D glyphs on a 2D map. (b) Helix glyphs for cyclic patterns.
Creation of a non-planar 3D slice through space-time. (a) Topological path. (b) Geometrical path. (c) Extruded slice.
Spatial-temporal visualization along a wall on a map.
Facets to be considered when visualizing graphs. (a) Structure (b) Attributes (c) Time (d) Space (e) Groups
Node-link diagram of flights connecting US airports.
Node-link diagram and corresponding matrix representation.
Graph patterns represented as matrices and node-link diagrams.
Differently ordered matrix representations of the same data. (a) Ordered by name.
Differently ordered matrix representations of the same data. (b) Ordered by frequency.
Differently ordered matrix representations of the same data. (c) Ordered by community.
Node-link representation compared to implicit representations. (a) Node-link. (b) Inclusion. (c) Overlap. (d) Adjacency.
Implicit visualizations of a classification hierarchy. (a) Squarified Treemap. (b) Information pyramids. (c) 3D sunburst.
NodeTrix visualization of a co-author network.
Spatial composition of graph facets in a single representation. (a) Juxtaposition. (b) Superimposition. (c) Nesting.
Map with tree layouts embedded into selected regions.
Three map layers visualize the data of three consecutive time steps. Spikes and lines indicate differences between the layers.
Multiple coordinated views for multi-faced graph visualization.
Stages of action forming the action cycle.
Conceptual separation across different models.
Spatial separation between the graphical user interface (right) and the visual representation in the main view (center).
Three-state model of graphical input.
Rubberband selection for marking multiple data elements. (a) Selection by inclusion.
Rubberband selection for marking multiple data elements. (b) Selection by intersection.
Four steps of selecting multiple trajectories using modifier keys.
Selecting segments based on their color, which represents speed. (a) How to select slow-speed segments?
Selecting segments based on their color, which represents speed. (b) Select via an interactive legend!
Selecting nodes based on their data attributes. (a) How to select high-degree nodes?
Selecting nodes based on their data attributes. (b) Select via slider handles!
Strategies for visual emphasis of relevant data and attenuation of less-relevant data.
Visual feedback for selections in visual representations of graphs. (a) Original visual representation.
Visual feedback for selections in visual representations of graphs. (b) Highlighting by encircling nodes.
Visual feedback for selections in visual representations of graphs. (c) Dimming nodes and edges.
Visual feedback for selections in visual representations of graphs. (d) Filtering nodes and edges.
Brushing a range (red) of an axis for binary and fuzzy selection. (a) Binary selection in the range.
Brushing a range (red) of an axis for binary and fuzzy selection. (b) Fuzzy selection beyond the range.
Brushing & linking in a multiple-views graph visualization.
Using V^= to scale relevant data to fit the display space.
Illustration of the conceptual model of zoomable interfaces.
Geometric zooming of a node-link visualization.
Semantically enhanced zooming of a node-link visualization.
A zoomable graph visualization and its controls.
Visual cues for pointing to off-screen data.
Bring & go with radar view and proxy nodes.
Viewports during an animated transition.
Snapshots of the viewport animation outlined in Figure 4.21.
A range slider controls the time period mapped to a spiral visualization of the daily average temperature for the city of Rostock.
Adjusting a time period at different input scales. (a) Regular range slider with global scale.
Adjusting a time period at different input scales. (b) A slider with increased precision is dynamically added to the range slider.
Integrated sliders for nD pan and zoom in the TimeWheel. (a) Plain non-interactive axes.
Integrated sliders for nD pan and zoom in the TimeWheel. (b) Axes with integrated sliders.
Integrated range slider for per-axis pan and zoom.
Schema of an interactive lens.
Model of a lens pipeline attached to a standard visualization.
Fundamental effects of lens functions. (a) Alteration.
Fundamental effects of lens functions. (b) Suppression.
Fundamental effects of lens functions. (c) Enrichment.
Lenses with different shapes and orientation. (a) Circular.
Lenses with different shapes and orientation. (b) Rectangular orientable.
Lenses with different shapes and orientation. (c) Content-adaptive shape.
Direct manipulation of lenses. (a) Move and resize.
Direct manipulation of lenses. (b) Adjust parameters.
Magnifying details in a map visualization with a fish-eye lens. (a) Regular map visualization.
Magnifying details in a map visualization with a fish-eye lens. (b) Details magnified with a fish-eye lens.
Graph lenses for exploring structural relationships. (a) Node-link diagram without lens.
Graph lenses for exploring structural relationships. (b) Local-edge lens.
Graph lenses for exploring structural relationships. (c) Bring-neighbors lens.
Graph lenses for exploring structural relationships. (d) Composite lens.
A lens to query temporal characteristics of movement data.
Orthogonal node-link diagram of a biological network.
Editing using the edit lens. (a) Place lens to insert.
Editing using the edit lens. (b) Adjust lens to update.
Editing using the edit lens. (c) Flick lens to delete.
Visual designs for comparison tasks.
Natural behavior of people comparing information on paper. (a) Side-by-side.
Natural behavior of people comparing information on paper. (b) Shine-through.
Natural behavior of people comparing information on paper. (c) Folding.
Creating sub-views for comparison. A red frame indicates where the left sub-view has been detached from the main view.
Overview of natural interaction techniques for visual comparison.
Information-rich, natural, and occlusion-free folding styles.
Relocating selected regions to form a ring for easier comparison. The map background has been desaturated for the purpose of illustration.
Visualizing future appointments with a SpiraClock.
Extended interaction with tangible views.
A circular tangible lens for magnification purposes.
Comparing matrix data with two tangible views.
Tangible views for different visualizations. (a) Parallel coordinates.
Tangible views for different visualizations. (b) Node-link diagram.
Tangible views for different visualizations. (c) Space-time cube.
Interacting by physical movements. (a) Zones for global control.
Interacting by physical movements. (b) Gaze plus lens for local control.
Procedure of determining visual density in parallel coordinates. (a) Binned axes. (b) Bin map. (c) Categorization. (d) Drawing.
User-selected data in red compared against general trends in green.
General procedure of bundling.
Visualization of dependencies in a software class hierarchy. (a) Conventional representation.
Visualization of dependencies in a software class hierarchy. (b) Hierarchical edge bundling (Holten, 2006)
Illustration of Furnas' DoI function. (a) Distances to focus node. (b) Adding node levels. (c) Extracted subtrees.
Five years of a dynamic co-author network extracted from DBLP.
Glyph design for representing collapsed subgraphs.
Feature specification with an interactive interface. (a) Specification of thresholds.
Feature specification with an interactive interface. (b) Formal feature definition.
Comparison of direct volume visualization of the particle concentration of one protein and ellipsoid-based visualization of features representing high concentrations of two different proteins. (a) Volume visualization.
Comparison of direct volume visualization of the particle concentration of one protein and ellipsoid-based visualization of features representing high concentrations of two different proteins. (b) Ellipsoid visualization.
Visualizing the temporal evolution of features. (a) Node-link diagram of the event graph.
Visualizing the temporal evolution of features. (b) Two features at two time steps.
Drawing thousands of trajectories of chaotic movements for multiple simulations leads to cluttered and indecipherable visual representations.
From entities (dot marks) to density map (gray-scale image) to regions (colored image).
2D movement reduced to 1D time series of feature values.
Feature visualization with overview and detail views.
Visualization of parameter settings, feature values, and detail information for selected parts of the data. (a) Parameter settings as gray-scale matrix; (b) Feature values over time as color-coded matrix; (c) Chart with selected time series; (d) Trajectory view with selected trajectory segments.
Visualizing the parameter dependency of average group size.
Visualization of the average distance to free proteins reveals the sweeping effect. (a) Density map. (b) Average protein-raft distance.
Visualization of a time series with more than 1.7 million time points, where each black pixel represents about 1,000 data points.
One and the same time series at two different scales.
Unifying the sample points of two successive scales by mapping and interpolation.
Computing the absolute value difference (AVD) and the slope sign difference (SSD) between two successive data scales.
Aggregation of data differences with maximum aggregation for the absolute value difference (AVD) and average aggregation for the slope sign difference (SSD) function.
Visualizing aggregated differences along with the actual data.
Time series plot of the simulation outcome and corresponding multi-scale difference bands of SSD and AVD with local color mode.
Studying the details of the middle peak-notch-spike pattern from Figure 5.28.
A decision tree classifying enterprises according to their sales.
Illustration of activity recognition based on parameter-dependent algorithms that learn from some ground truth.
Visualization of parameter-dependent classification outcomes for activity recognition. (a) Parameter configurations; (b) Recognized activities; (c) Color legend; (d) Ground truth; (e) and (f) Stacked histograms with aggregated information.
Highlighting the incorrectly classified time steps in red.
Illustration of clustering strategies. (a) Input data. (b) K-means. (c) Ward's method. (d) STING. (e) DBSCAN. (f) Dendrogram.
Comparison of clusters generated with hierarchical clustering, k-means, and affinity propagation. (a) Hierarchical clustering. (b) K-means clustering. (c) Affinity propagation.
Applying hybrid SOM-based clustering to sort rows in a table lens visualization. (a) Unordered rows before clustering.
Applying hybrid SOM-based clustering to sort rows in a table lens visualization. (b) Ordered rows after clustering.
Two-step procedure of clustering nodes based on their attributes. First, nodes with similar attribute behavior are grouped. Second, groups are refined based on connected components. (a) Initial grouping. (b) Refined clusters.
Structure-based clustering. (a) Initial set of states and transitions based on the sequence of graphs Gi in DG; (b) Hierarchical grouping of states and transitions based on similar structures.
Example of a state-transition graph characterizing an underlying dynamic graph.
Analyzing a wireless network supported by structure-based clustering. (a) State-transition graph; (b) Average link quality of selected state; (c) Representative graph structure of selected state.
Reducing dimensionality with principal component analysis. (a) Original data space. (b) Principal component space. (c) Reduced space.
Overview of automatic computational methods to support interactive visual data analysis by reducing the complexity of the data and their visual representations.
Graphical interface for creating and controlling multi-display visual analysis presentations, including content pool (top), logical presentation structure (middle), and preview (bottom).
Basic two-step procedure of the automatic view layout.
Changing the content of views by launching visualization software.
Graphical interface for analysis coordination and meta analysis, including filtering support (top), analysis history graph (middle), and timeline with undo and redo buttons (bottom).
A knowledge gap exists when target or path are unknown.
Adapted variant of van Wijk's model of visualization. Artifacts as boxes: data [D], specifications [S], visualization images [I], and user knowledge [K]. Functions as circles: analytic and visual transformation (T), perception and cognition (P), and interactive exploration (E).
Conceptual model of guided interactive visual data analysis. *Added artifacts and functions: domain conventions and models [D*], history and provenance [H*], visual cues [C*], and options and alternatives [O*], and guidance generation (G*).
Navigation recommendations for graph visualization.
Tailored domain model as the basis for user guidance.
Incremental processes highlighted in van Wijk's model of visualization.
Extended notation for operators and transitons for progressive visual data analysis.
Simple example of a progressive transformation pipeline.
A multi-threading architecture for progressive visual data analysis.
Illustration of asynchronous processing threads operating on data chunks stored in priority queues.
Comparison of single-thread and multi-thread solutions. (a) Regular single-thread solution. (b) Progressive multi-thread solution.
The three typical scenarios of progressive visual data analysis: progressive data processing, progressive visualization, and progressive display.
Visualization of progressively processed data chunks of car crashes from a database with more than 370,000 entries. (a) Regular progression of chunks.
Visualization of progressively processed data chunks of car crashes from a database with more than 370,000 entries. (b) Prioritized progression of chunks.
Progressive force-directed layout of a social network with 747 nodes and 60,050 edges.
Progressive visualization of a climate network with about 6,816 nodes and 232,940 edges.
Device-dependent progressive transmission of a treemap visualization image.
Progressive display for a dynamically defined region of interest. (a) Global view. (b) Overview+detail view. (c) Focus+context view.