23CSE532 Data Visualization Techniques

|| Jai Sri Gurudev |

Sri AdichunchanagiriShikshana Trust (R)

SJB INSTITUTE OF TECHNOLOGY

Study Material
Module 2

Subject Name: Data Visualization Techniques

Subject Code: 23CSE532
Semester: V

By

Faculties Name:

Dr. Prakruthi M K, Mrs. Vinutha K,

Assoc. Professor, Asst. Professor,
Dept. of CSE, Dept. of CSE,
SJBIT. SJBIT.

Department of Computer Science & Engineering

Aca. Year: Even Sem / 2025-26

1
Dept. of CSE,SJBIT
23CSE532 Data Visualization Techniques

Module-2: Visualization Foundations and Techniques

Visualization Foundations: The Visualization Process in Detail, Semiology of Graphical

Symbols, The Eight Visual Variables, Taxonomies - Taxonomy of Visualization Goals, Data
Type by Task Taxonomy, Keim Taxonomy Visualization Techniques for Spatial Data: One-
Dimensional Data, Two-Dimensional Data, Three-Dimensional Data

Textbook 1: Chapter 4 and 5

Self-Learning: Study advanced taxonomy frameworks for visualization beyond Keim’s model (e.g.,
Munzner’s Nested Model).

Steps in the Visualization Process

(Figure 4.1 is the reference pipeline)

2
Dept. of CSE,SJBIT
23CSE532 Data Visualization Techniques

1. Data Preprocessing and Transformation

o Start with raw data → make it usable.
o Steps:
 Map to fundamental datatypes (numbers, strings, etc.).
 Handle missing values (interpolation).
 Handle errors in input.
 Handle large datasets (sampling, filtering, aggregation, partitioning).
o Data may come from simulation or sampling.

3
Dept. of CSE,SJBIT
23CSE532 Data Visualization Techniques

2. Mapping for Visualization

o Choose how to represent cleaned data visually (geometry, color, sound, etc.).
o Danger: Wrong mappings → misleading or nonsense visualization.
o Example:
 Bad: Using bar charts incorrectly (Figure 4.2a).
 Better: Using scatterplots (Figure 4.2b).
o Key Influences:
 Expressiveness → Does it present all and only the information?
 Effectiveness → Is it easy to interpret and cost-efficient?
3. Rendering Transformations
o Final step: map geometry → images using graphics APIs (OpenGL, Processing, Java,
Flex, etc.).
o Includes:
 Viewing parameters.
 Shading techniques (for 3D).
 Device transformations (screen, printer).
o Highly dependent on the graphics library.

Measures of Visualization
4
Dept. of CSE,SJBIT
23CSE532 Data Visualization Techniques

Example (Figures 4.3a & 4.3b)

 Both are expressive (all and only relevant car data shown).
 But effectiveness differs:
5
Dept. of CSE,SJBIT
23CSE532 Data Visualization Techniques

o One visualization makes it quicker to answer "Which car has best mileage?"
o Another makes it quicker to answer "Which car under $11,000 has best mileage?"

✅ Key takeaway:
The visualization pipeline = Preprocessing → Mapping → Rendering, with user interaction possible
at every stage.
Quality of visualization = judged by expressiveness (what is shown) and effectiveness (how well it
can be interpreted).

4.2 Semiology of Graphical Symbols

Definition

 Graphical Symbol → A visual object used in a visualization (e.g., arrows, labels, shapes,
icons, points).
 Semiology → The science/study of how graphical symbols convey meaning and how they
are perceived by humans.
 Scope → Any construction in the Euclidean plane (diagrams, networks, maps, plots, charts)
is composed of graphical symbols.
 Goal → To make data visible, readable, and interpretable by exploiting three key features:
o Similarity (same-looking symbols = similar data).
o Ordering (symbol arrangement reflects data order).
o Proportionality (symbol size/length/area corresponds to data magnitude).

4.2.1 Symbols and Visualizations

6
Dept. of CSE,SJBIT
23CSE532 Data Visualization Techniques

 Simple Symbol (Figure 4.4a)

o Example: Yield sign.

o Recognized immediately (preattentive recognition).

o One-step interpretation: Direct association with meaning.

 Complex Symbol (Figure 4.4b)

o Requires multi-step attention.

o Steps:

1. Recognize main elements of the image.

2. Identify relationships among these elements.

3. Extract patterns (groups, clusters, features).

4. Identify “interesting” cases (special features, outliers).

 Key Point:

o Without external identification (labels, axes, legends), graphics are unusable.

o Meaningful graphics must:

 Have clear and interpretable axes (x, y, z).

 Be directly understandable.

7
Dept. of CSE,SJBIT
23CSE532 Data Visualization Techniques

Pattern Discovery Rules

 Any visual pattern must reflect a real data pattern.

 Otherwise → misleading visualization.

 Rules of correlation:

o Visual similarity ⇔ Similarity in data.

o Visual order ⇔ Order in data.

 Misalignment = artifact of representation.

4.2.2 Features of Graphics

1. Dimensionality of Graphics

o Graphics typically use three variables (x, y, z).

o Example: Figure 4.5 (matrix with circles of different sizes).

 X = row, Y = column, Z = size of circle.

o 3D view (height) ⇔ 2D view (size).

o Patterns along diagonals or groups immediately pop out.

8
Dept. of CSE,SJBIT
23CSE532 Data Visualization Techniques

2. Perception of Groups

o Example: Figure 4.6 (tree structure).

o Human eye perceives whole structures first (tree), not individual leaves.

o Texture and distributions of symbols may themselves form patterns.

o Eye naturally detects groups, distributions, and dominant features.

Rules of a Graphic
1. Aim: Discover groups/orders in (x,y), formed on z-values.

2. (x,y,z) construction always enables discovery of groups.

3. Permutations & classifications help reveal upper-level info.

4. Graphics with >3 factors (beyond x,y,z) may destroy unity.

5. Every graphic must be readable and understandable by humans.

9
Dept. of CSE,SJBIT
23CSE532 Data Visualization Techniques

Analysis of a Graphic
 Process of interpretation:

1. Preattentive stage → perceive groups.

2. Cognitive stage → characterize groups.

3. Detailed stage → find exceptions, relationships between groups.

 Supporting analysis:

o Computations (e.g., clustering) can guide perception by highlighting groups.

✅ Key Takeaway:
 Semiology = grammar of graphics.

 Good symbols = direct, interpretable, match data patterns.

 Bad symbols = misleading, complex, or add false patterns.

 Always align visual similarity & order with data similarity & order.

The Eight Visual Variables

 Visualization is the process of mapping data to graphics for effective communication.

 The basic units of graphics are marks (points, lines, areas, volumes, etc.).

 Simply plotting marks without structure leads to clutter.

 Spatial arrangement and visual properties of marks are what make data meaningful.

 To make a visualization effective, we use eight visual variables (Bertin’s framework).

1. Position (Most important variable)

 Refers to placement of marks in 1D, 2D, or 3D space.

 Human vision is highly sensitive to spatial differences → easiest to interpret.

 Key role: Shows distribution, trends, clusters, and relationships between variables.

 Example: Scatterplot (Car dataset → min price vs. max price).

10
Dept. of CSE,SJBIT
23CSE532 Data Visualization Techniques

 Good use: Spread out marks to avoid overlap.

 Scales:

o Linear scale → direct proportional mapping.

o Logarithmic scale → compresses exponential data.

 Projection: Used in maps and dimensionality reduction (e.g., cartography).

2. Mark (Shape)
 Marks = graphic primitives that represent data (points, lines, polygons, symbols).

 Shape differentiates categories.

 Must ensure distinct and balanced shapes (similar area & complexity) to avoid bias.

 Shapes that leverage perception (like T, L, +, –) are effective.

 Example: Different car types shown as circle, triangle, or square in a scatterplot.

3. Size (Length, Area, Volume)

 Controls how big or small a mark appears.

 Works well for interval and continuous variables (gradual variation).

 Categorical use is limited → hard to distinguish small size differences.

 Works best with points/lines, less with areas/volumes.

 Example: Bubble chart (Car engine size vs. fuel tank capacity, bubble size = price).

 Note: Perception of area/volume is less accurate than length.

4. Brightness (Luminance)
 Refers to lightness/darkness of a mark.

 Useful for relative comparisons in large datasets.

 Humans cannot distinguish too many brightness levels → best used with limited steps.

11
Dept. of CSE,SJBIT
23CSE532 Data Visualization Techniques

 Should use perceptually linear scales (equal perceptual differences).

 Example: Darker dots = wider cars, lighter dots = narrow cars.

5. Color (Hue and Saturation)

 Hue: actual color (red, green, blue, etc.).

 Saturation: purity/intensity of the color.

 Color is powerful but must be used carefully (perception and cultural meanings matter).

 Usage:

o Continuous data → colormaps (gradual blending of hues).

o Categorical data → distinct colors.

 Examples:

o Blue → short car, Yellow → long car (Car length visualization).

o Colormaps: Grayscale, rainbow, heated, diverging blue–yellow.

6. Orientation (Direction)
 Rotation of a mark to show variation.

 Works only with asymmetric marks (triangles, arrows, bars).

 Human vision quickly detects orientation changes (preattentive).

 Example: Triangular marks rotated clockwise for increasing car price.

7. Texture
 Combination of repeated patterns (dots, stripes, grids).

 Differentiates surfaces, categories, or regions.

 Works well for polygonal areas, surfaces, or 3D objects.

 Types: geometric texture (physical ridges) or color texture (visual variation).

12
Dept. of CSE,SJBIT
23CSE532 Data Visualization Techniques

 Example: Different textures (dotted/striped) for car categories in horsepower vs. wheelbase
plot.

8. Motion
 Dynamic change of a variable over time.

 Can be applied to any other variable (position, size, brightness, etc.).

 Attracts attention instantly, highlights trends/outliers.

 Types:

o Speed (fast/slow movement).

o Direction (up, down, diagonal, circular).

o Flashing/blinking (changing opacity or brightness).

 Example: Animated scatterplot showing how car prices changed over years.

Supporting Elements
 Scales: Linear & Logarithmic scaling make variables interpretable.

 Projection: Mapping higher dimensions to 2D (e.g., maps, PCA).

 Axes: Provide context (tick marks, labels, titles). Essential for interpretation.

Why Important?
 Helps answer key questions:

o Where do most data values fall?

o Are there clusters or outliers?

o Do variables show correlation or independence?

 Visualization must consider human perception limits (Gestalt laws, preattentive vision, etc.).

13
Dept. of CSE,SJBIT
23CSE532 Data Visualization Techniques

Exam-Oriented Pointers
 Position + Mark = Foundation → Without them, no visualization exists.

 Other six variables (size, brightness, color, orientation, texture, motion) = enhancements.

 Common exam questions:

o Define and explain the eight visual variables with examples.

o Why is position the most effective variable?

o Distinguish brightness vs. color as visual variables.

o Explain the role of size in representing continuous vs. categorical data.

o Discuss the importance of texture and motion in visualization.

Effects of Visual Variables

Visual variables do not just encode data—they also influence how humans perceive and interpret
the visualization. Their effects can be grouped into categories:

1. Selective Visual Variables

 Purpose: Instantly separate data into distinct groups. Useful for nominal (categorical)
values.

 Effective variables:

o Size (length, area, volume)

o Brightness

o Texture (sample pattern)

o Color (primary colors, with brightness variation)

o Direction/orientation

 Example: Using red vs. blue vs. green dots to separate car types.

2. Associative Visual Variables

14
Dept. of CSE,SJBIT
23CSE532 Data Visualization Techniques

 Purpose: Show that all groups are equally visible (no group dominates). Useful when all
categories are equally important.

 Effective variables:

o Texture (patterns like dots, stripes)

o Color (balanced hues)

o Orientation (different rotations)

o Shape (different symbols)

 Example: Pie chart with equal emphasis on all slices, each with a different color/shape.

3. Ordinal Visual Variables

 Purpose: Naturally show order or ranking (best for ordinal or quantitative data).

 Effective variables:

o Texture (denser → lighter patterns)

o Size (smaller → larger marks)

o Brightness (light → dark)

 Example: Bar chart where taller bars = higher values, or brightness scale from light (low) to
dark (high).

4. Proportional Visual Variables

 Purpose: Provide a direct sense of relative magnitude. Useful for quantitative
comparisons.

 Effective variables:

o Size (length/area/volume proportional to data value)

o Orientation/direction (e.g., arrow length & direction showing wind force)

o Brightness (darker = more, lighter = less)

15
Dept. of CSE,SJBIT
23CSE532 Data Visualization Techniques

 Example: Bubble chart where bubble size ∝ population.

5. Separating Visual Variables

 Purpose: Clearly distinguish one group of data from the rest (foreground vs. background).

 Effective variables:

o Texture

o Color

o Orientation

o Shape

 Example: Highlighting one cluster in a scatterplot with a unique color/shape, while others
remain muted.

Summary Table: Effects of Visual Variables

Category Purpose Effective Visual Variables Example
Instantly divide into groups Size, Brightness, Texture, Color, Red vs. blue dots for
Selective
(nominal) Orientation categories
Texture, Color, Orientation, Pie chart with different
Associative Equal visibility of groups
Shape hues
Show ranking/order (ordinal, Darker shades = higher
Ordinal Texture, Size, Brightness
quant.) values
Proportional Direct sense of magnitude Size, Orientation, Brightness Bubble size = population
Distinguish subset from Texture, Color, Orientation, Highlighted cluster in
Separating
others Shape scatterplot

Taxonomies - Taxonomy of Visualization Goals, Data Type by Task Taxonomy, Keim

Taxonomy

16
Dept. of CSE,SJBIT
23CSE532 Data Visualization Techniques

4.5 Taxonomies in Visualization

What is a Taxonomy?
 A taxonomy is a structured classification system, often hierarchical, that organizes similar
entities and explains their relationships.

 In data visualization, taxonomies help us:

o Understand different types of data.

o Organize visualization techniques systematically.

o Classify tasks/goals that visualization supports.

o Define interaction methods for effective exploration.

 Purpose: Helps designers and researchers choose the right visualization approach depending
on data characteristics and user requirements.

4.5.1 Taxonomy of Visualization Goals (Keller & Keller)

Background
 Proposed by Keller and Keller in their book Visual Cues.

 They created a classification based on:

1. The type of data being analyzed.

2. The user’s tasks or goals.

 This framework was applied to more than 100 visualization techniques, making it one of the
early systematic approaches to categorizing visualization.

(A) Data Types

1. Scalar (scalar fields) – Single numeric values (e.g., temperature, pressure, elevation).

2. Nominal – Categorical data (e.g., gender, blood group, product type).

3. Direction (direction fields) – Vector-based data (e.g., wind directions, flow fields).

17
Dept. of CSE,SJBIT
23CSE532 Data Visualization Techniques

4. Shape – Geometric information (e.g., 3D model outlines, anatomical shapes).

5. Position – Spatial coordinates (e.g., locations on a map).

6. SERO (Spatially Extended Region or Object) – Larger structures such as maps,

geographic areas, or 3D volumetric objects.

(B) User Tasks

Keller & Keller identified 9 major tasks that a user might perform while exploring data:
1. Identify – Recognize characteristics that make an object unique.

o Example: Identifying a cancerous cell in medical imaging.

2. Locate – Find position of an object (absolute or relative).

o Example: Locating a country on a world map.

3. Distinguish – Recognize something as different from others (without full identification).

o Example: Spotting outliers in a scatter plot.

4. Categorize – Place objects into classes or groups.

o Example: Grouping customers into “high spenders” vs. “low spenders.”

5. Cluster – Group together similar objects (more data-driven than categorization).

o Example: Clustering genes with similar expression patterns.

6. Rank – Order items based on some property.

o Example: Ranking universities based on placement percentage.

7. Compare – Observe similarities and differences.

o Example: Comparing GDP growth between two countries.

8. Associate – Link objects in a relationship.

o Example: Associating weather patterns with crop yield.

9. Correlate – Establish a direct or causal relationship.

18
Dept. of CSE,SJBIT
23CSE532 Data Visualization Techniques

o Example: Correlating smoking with lung cancer rates.

Importance of Keller & Keller’s Taxonomy

 Provides a task-oriented classification of visualization techniques.

 Helps designers choose visualization methods by combining data type and task type.

 Example: If the data is scalar and the task is compare, you might use a bar chart or line
plot.

4.5.2 Data Type by Task Taxonomy (Shneiderman)

Background
 Proposed by Ben Shneiderman, a pioneer in information visualization.

 His taxonomy is broader and focused more on interactive exploration of data.

 Basis: Visualization should support analysts in their behavioral workflow of knowledge

discovery.

(A) Data Types

1. One-dimensional linear – Sequences (e.g., text lines, lists, stock prices).

2. Two-dimensional map – Geographical or layout-based data (e.g., city maps, floor plans).

3. Three-dimensional world – 3D physical models (e.g., molecular structures, CAD models).

4. Temporal – Time-series data (e.g., weather over days, stock prices over years).

5. Multidimensional – Data with many attributes (e.g., spreadsheets, customer records).

6. Tree – Hierarchical structures (e.g., file systems, organizational charts).

7. Network – Graphs with nodes and links (e.g., social networks, computer networks).

(B) User Tasks

Shneiderman identified 7 tasks, which later evolved into his famous Information-Seeking Mantra:
“Overview first, zoom and filter, then details-on-demand.”
19
Dept. of CSE,SJBIT
23CSE532 Data Visualization Techniques

1. Overview – Gain an overall picture of the dataset.

o Example: Viewing the entire metro map before selecting a station.

2. Zoom – Focus on items of interest by enlarging specific regions.

o Example: Zooming into a city in Google Maps.

3. Filter – Remove irrelevant or uninteresting data items.

o Example: Filtering job listings by salary or location.

4. Details-on-demand – Get additional details when required.

o Example: Hovering on a stock point to see exact value.

5. Relate – Explore relationships between objects.

o Example: Clicking a person in LinkedIn to see their connections.

6. History – Keep track of user actions for undo/redo or replay.

o Example: Browser history or workflow log in Tableau.

7. Extract – Export selected data for use in another context.

o Example: Downloading filtered survey results into Excel.

Importance of Shneiderman’s Taxonomy

 More interaction-centric than Keller & Keller’s.

 Provides a workflow for data exploration.

 Suitable for designing visual analytics systems.

 Still widely quoted in visualization research and tools.

Comparison: Keller & Keller vs. Shneiderman

Aspect Keller & Keller Shneiderman
Focus Classifying visualization goals based on data + Classifying data + defining

20
Dept. of CSE,SJBIT
23CSE532 Data Visualization Techniques

Aspect Keller & Keller Shneiderman

tasks interactive tasks
Scalar, Nominal, Direction, Shape, Position, 1D, 2D, 3D, Temporal,
Data Types
SERO Multidimensional, Tree, Network
Identify, Locate, Distinguish, Categorize, Overview, Zoom, Filter, Details,
Tasks
Cluster, Rank, Compare, Associate, Correlate Relate, History, Extract
Orientation Analytical, static classification Interactive, dynamic exploration
Key Early systematic framework for mapping “Information-Seeking Mantra”
Contribution techniques guiding user workflow

Key Takeaways
Keller & Keller → Task-driven taxonomy (focus on what users want to do).
 Shneiderman → Interaction-driven taxonomy (focus on how users explore data).

 Both are complementary → one organizes goals, the other organizes interaction methods.

Got it ✅ You want me to elaborate this section on Visualization Techniques for Spatial Data (esp.
One-Dimensional Data) so that it is more exam-ready, with deeper explanations, clear flow, and
some intuitive examples. Let me expand it for you:

Visualization Techniques for Spatial Data (Elaborated)

Spatial data visualization deals with data that has implicit or explicit spatial (location-based) or
spatio-temporal (location + time) attributes. This is the basis of scientific visualization, where the
physical or simulated world is represented visually for better analysis and understanding.
Why spatial visualization is natural
Our visual perception system is continuously processing spatial information (depth, motion, scale,
texture). Hence, mapping scientific data to spatial representations feels intuitive. Computers extend
this ability with more flexibility:
 Unlike real-world perception (continuous, 3D, high resolution), computer visualization is
usually 2D, discrete, and screen-limited.

 On the screen, we can:

21
Dept. of CSE,SJBIT
23CSE532 Data Visualization Techniques

o Zoom into arbitrary scales (microscopic to astronomical).

o Adjust contrast, lighting, resolution, and density dynamically.

o Navigate through otherwise inaccessible environments (e.g., inside a volcano model).

o Interactively filter/remove data to reduce clutter or add missing context.

Thus, visualization transforms data into a form aligned with how humans already interpret the world.

Key Design Step

In visualization of spatial data:
1. Map spatial attributes of the data to screen space (location on X, Y, or Z axis).

o Requires geometric transformations: scaling, translation, rotation, shearing,

projection.

2. Map non-spatial attributes (measurements, properties, categories) to graphical attributes

such as:

o Color

o Size

o Shape

o Texture

5.1 One-Dimensional (1D) Data

1D spatial data arises when measurements are taken along a single path in space.
Examples:
 Drill-hole samples: ore/mineral concentration vs. depth.

 Temperature along a pipeline.

 Elevation profile along a road.

This type of data is often used as a probe to study higher-dimensional structures.

22
Dept. of CSE,SJBIT
23CSE532 Data Visualization Techniques

Visualization Techniques for 1D Data

Case 1: Univariate 1D Data (only one value per sample)
Two main methods:
1. Line Graph

o X-axis → spatial dimension (e.g., depth, distance).

o Y-axis → measured value.

o Displays trends, peaks, periodicity.

o Example: elevation profile of a hiking trail.

2. Color Bar (1D color mapping)

o X-axis → spatial dimension.

o Color intensity or hue → data value.

o Example: showing temperature along a road with color changes (blue = cool, red =
hot).

Both techniques require scaling raw data values to the screen coordinate or color range.

23
Dept. of CSE,SJBIT
23CSE532 Data Visualization Techniques

Case 2: Bar Graph

 Each data point → rectangle (bar).

 Height proportional to value.

 Useful when showing discrete samples.

 Can be colored for categorical variables.

Case 3: Multivariate 1D Data

When multiple variables are recorded along the same spatial path (e.g., mineral content + moisture +
density in drilling):
1. Juxtaposition – show multiple separate line/bar graphs (stacked).

24
Dept. of CSE,SJBIT
23CSE532 Data Visualization Techniques

o Best when variables have different scales.

2. Superimposition – overlay multiple variables on the same graph.

o Requires different line styles, colors, or markers.

o Risk: clutter, confusion.

3. Stacked Bar Graphs – bars subdivided into components (easier interpretation).

Additional Enhancements
 Logarithmic scaling: when data values span large ranges.

 Downsampling or averaging: when too many points exceed screen resolution.

 Interactive zooming: explore smaller subsets with detail.

 Color encoding: additional dimension of data.

Summary of 1D Visualization
 Line Graphs → trends and variations.

 Bar Graphs → discrete samples, comparison.

 Color Bars → continuous distribution with easy perception.

 Multivariate Methods → overlay or stacking for multiple variables.

1D techniques form the foundation for higher-dimensional visualization (2D/3D data often reduced
to 1D slices or profiles).

Keim’s Taxonomy of Visualization Systems

Background
 Proposed by Daniel A. Keim.

 Focus: A three-dimensional classification scheme for visualization systems.

 Dimensions used:

1. Data Types
25
Dept. of CSE,SJBIT
23CSE532 Data Visualization Techniques

2. Visualization Techniques

3. Interaction & Distortion Techniques

👉 This makes Keim’s taxonomy more comprehensive than Keller & Keller or Shneiderman, because
it includes visualization techniques explicitly, along with tasks/interactions.

1. Classification of Data Types

Keim defines six categories of data types that visualization systems deal with:
1. One-dimensional data

o Sequential data.

o Examples: Temporal series (stock prices, ECG signals, weather data), text streams,
news articles.

2. Two-dimensional data

o Structured in a 2D plane.

o Examples: Maps, charts, floor plans, newspaper layouts.

3. Multidimensional data

o Data with multiple attributes (n-dimensional).

o Examples: Spreadsheets, relational databases, data warehouse tables.

4. Text and hypertext data

o Unstructured or semi-structured textual data.

o Examples: Web documents, research papers, blogs, online news articles.

5. Hierarchies and graphs

o Data represented in tree or network form.

o Examples: Organization charts, file systems, telephone/network traffic, system

dynamics models.

6. Algorithms and software data

26
Dept. of CSE,SJBIT
23CSE532 Data Visualization Techniques

o Computational data related to software execution.

o Examples: Execution traces, program call graphs, memory dumps.

2. Classification of Visualization Techniques

Keim classified five classes of visualization techniques:
1. Standard 2D/3D Displays

o Basic plots and charts.

o Examples: Scatter plots, bar charts, line graphs, 3D surface plots.

2. Geometrically-Transformed Displays

o Transform data geometrically into new visual forms.

o Examples:

 Scatterplot matrices

 Parallel coordinates

 Projection pursuit

 Hyperslice

 Landscapes

3. Iconic Displays

o Represent data attributes using icons.

o Examples:

 Chernoff faces (facial features represent attributes)

 Star glyphs

 Stick figures

 Tilebars

4. Dense Pixel Displays

27
Dept. of CSE,SJBIT
23CSE532 Data Visualization Techniques

o Map data values directly to colored pixels, producing dense visual encodings.

o Examples:

 Recursive pattern

 Circle segments

 Pixel-based sketches

5. Stacked Displays

o Partition screen space hierarchically to represent multidimensional data.

o Examples:

 Dimensional stacking

 Treemaps

 Cone trees

 Worlds-within-worlds

3. Classification of Interaction & Distortion Techniques

Keim also recognized five types of interaction techniques, which allow users to manipulate and
explore data more effectively:
1. Dynamic Projection

o Change projections of multidimensional data dynamically.

o Examples: Grand Tour System, XGobi, XLispStat.

2. Interactive Filtering

o Filter unwanted data to focus on relevant subsets.

o Examples: Dynamic queries, Magic Lenses, InfoCrystal, Polaris.

3. Interactive Zooming

o Zoom in/out for detail or overview.

28
Dept. of CSE,SJBIT
23CSE532 Data Visualization Techniques

o Examples: TableLens, PAD++, IVEE/Spotfire, DataSpace, MGV.

4. Interactive Distortion

o Distort data views to emphasize focus regions while preserving context.

o Examples: Fisheye views, Perspective wall, Hyperbolic visualization, Bifocal

displays.

5. Interactive Linking and Brushing

o Link multiple coordinated views; brushing highlights selected items across views.

o Examples: Parallel coordinates + scatter plots, pixel displays + maps, tools like
XmdvTool, S-Plus, Polaris.

Summary of Keim’s Taxonomy

 Data types: Defines what kind of data is being visualized.

 Visualization techniques: Defines how the data is represented visually.

 Interaction techniques: Defines how the user explores/manipulates the data.

Comparison with Other Taxonomies

 Keller & Keller → Focused on data types + tasks.

 Shneiderman → Focused on data types + interactive exploration tasks (overview–zoom–

filter–details mantra).

 Keim → More comprehensive: includes data types, visualization techniques, and

interaction methods in one framework.

Visualization Techniques for Spatial Data: One-Dimensional Data, Two-Dimensional Data,

Three-Dimensional Data

5.2 Visualization Techniques for Two-Dimensional (2D) Data (Elaborated)

When data has two spatial dimensions (X and Y), visualization becomes more powerful because
29
Dept. of CSE,SJBIT
23CSE532 Data Visualization Techniques

the screen itself is two-dimensional, so we can directly map the data space to the display space.
The visualization techniques vary depending on how the data is represented and the type of
information we want to extract.

Main Visualization Techniques for 2D Data

1. Image Representation
 Concept: Each data value is mapped to a color or intensity at a specific pixel location.

 Interpolation is applied between sample points to fill in missing pixels and make a smooth
image.

 Example: Medical imaging (CT scan, MRI), thermal imaging.

 Advantage: Easy and intuitive for humans to interpret continuous variation.

2. Rubber Sheet (Surface Representation)

 Concept: The data is treated like a flexible sheet stretched over a plane.

 Each (X, Y) coordinate is mapped to a height (Z-value), forming a 3D surface.

 Regular/Irregular Data: Works even if data points are irregularly spaced (triangulation fills
the gaps).

 Example: Terrain elevation maps, ocean floor mapping.

 Advantage: Natural way to visualize elevation or intensity distributions.

 Disadvantage: Can suffer from occlusion if surface folds back.

3. Cityscape
 Concept: Data is visualized as 3D objects (usually bars/boxes) on a flat plane.

 Attributes like height and color of bars represent data values.

 Example: Air traffic density across the U.S., population density per city block.

 Advantage: Very effective for discrete distributions.

 Limitation: Visual clutter for large datasets.

30
Dept. of CSE,SJBIT
23CSE532 Data Visualization Techniques

4. Scatterplot
 Concept: Each data point is represented by a marker (circle, dot, symbol).

 Marker color, shape, or size encodes the data value.

 No interpolation is performed (unlike images).

 Example: Plotting earthquake magnitudes by location, disease cases on a map.

 Advantage: Preserves actual data without smoothing.

 Limitation: Can become dense and unreadable for very large datasets.

5. Maps
 Concept: Represents linear, area, and point features.

o Linear features: Roads, rivers (drawn as connected line segments).

o Area features: Lakes, country borders (closed polygons, filled with textures or
colors).

o Point features: Schools, utility poles, hospitals (symbols).

 Labels are added for readability but require careful placement (cartography problem).

 Example: Google Maps, GIS maps.

 Advantage: Rich and intuitive.

 Limitation: Overlaps and clutter can reduce readability.

6. Contour (Isovalue) Maps

 Concept: Extract boundaries of equal value (isovalue).

 A contour line marks locations where the measured variable is constant.

 Computation:

o For each rectangle of 4 data points → determine where value crosses threshold.

31
Dept. of CSE,SJBIT
23CSE532 Data Visualization Techniques

o Use interpolation to find crossing points.

 Examples:

o Topographic maps (equal elevation lines).

o Weather maps (isotherms = equal temperature, isobars = equal pressure).

 Enhancements: Multiple contours with different colors, line thickness, or labels.

 Advantage: Very compact representation of continuous variation.

Multivariate 2D Data Visualization

When each (X, Y) point has multiple variables:
 Juxtaposition (Side-by-side visualization):

o Stack multiple images/surfaces in 3D (like MRI slices).

o Limitation: Occlusion makes it hard to compare.

 Superimposition (Overlay):

o Combine multiple variables on same visualization.

o Example: A cityscape where each block has stacked colored bars for different
variables.

o Limitation: Works only for small number of variables (<10), otherwise unreadable.

 Maps (layering):

o Multiple layers of roads, rivers, schools, vegetation overlaid.

o Problem: Too much overlap → clutter.

o Solution: Minor adjustments or selective hiding of layers.

If too many variables → switch to non-spatial multivariate visualization techniques (scatterplot

matrices, parallel coordinates, etc., in later chapters).

5.2.1 Probing 2D Data (Subsetting & Projection)

32
Dept. of CSE,SJBIT
23CSE532 Data Visualization Techniques

Sometimes instead of visualizing the full 2D dataset, we extract 1D information from it. Methods
include:
a) Frequency Histograms
 Summarize data values into ranges (bins).

 Plot frequency counts as a bar graph.

 Challenges: Choosing bin size and boundaries — too few bins hides detail, too many bins
show noise.

 Example: Pixel intensity histogram in image processing.

b) Row and Column Aggregations

 Summarize rows or columns using descriptors (sum, mean, median, max, standard
deviation).

 Produces 1D data which can be shown alongside the image.

 Example: Row sums in DNA microarray images to detect active gene regions.

c) Linear Probes
 Imagine drawing a line through a 2D dataset.

 Along that line, sample values → convert to 1D profile.

 Mathematical Basis:

o A line segment between points P1 and P2 can be represented as:

P(t)=P1+t(P2−P1),0≤t≤1P(t) = P1 + t(P2 - P1), \quad 0 \leq t \leq 1

o Interpolation gives values at intermediate points.

 Example: Cross-section of an MRI scan, profile of elevation along a hiking path.

Summary of 2D Techniques
 Images → continuous fields, interpolation.

 Rubber sheets → surface visualization.

33
Dept. of CSE,SJBIT
23CSE532 Data Visualization Techniques

 Cityscapes → bar-like 3D metaphors.

 Scatterplots → discrete data distribution.

 Maps → real-world feature representation.

 Contours → boundary extraction.

 Probes (histograms, row/column, linear) → simplification into 1D for analysis.

✨ 2D visualization is the backbone of spatial data analysis, combining raw representation

(images), abstract forms (cityscapes, scatterplots), real-world mapping (maps, contours), and probing
techniques to simplify exploration.

Three-Dimensional Data Visualization

Just like 2D data, 3D spatial data can be:
 Discrete samples of a continuous phenomenon (e.g., MRI scans, airflow measurements).

 Geometric structures defined by vertices, edges, and faces (e.g., CAD models, meshes).

Most real-world scientific and engineering visualizations (like stress on a machine part or airflow
around an aircraft wing) combine both types.

5.3.1 Visualizing Explicit Surfaces

Explicit surfaces are directly defined in geometry. Two main methods:
1. Using Vertex, Edge, and Face Lists
 Surfaces are represented by:

o Vertices → 3D coordinates (x, y, z).

o Edges → pairs of vertices.

o Faces → polygons formed by connecting edges.

 Example: Unit Cube representation uses 8 vertices, 12 edges, 6 faces.

34
Dept. of CSE,SJBIT
23CSE532 Data Visualization Techniques

 Orientation consistency (clockwise/counter-clockwise) is critical to compute surface

normals correctly (for shading & lighting).

2. Parametric Equations
 Surfaces defined using equations for x, y, z with parameters (u, v, θ, etc.).

 Example: Cylinder aligned to y-axis

x=cos⁡θ, z=sin⁡θ, 0≤θ≤2π, 0≤h≤1x = \cos \theta,\; z = \sin \theta,\; 0 \leq \theta \leq 2\pi,\; 0 \leq h \leq
1
 Step size of parameters controls smoothness (finer steps → smoother surfaces).

 Widely used in curves & splines (B-splines, Bézier surfaces).

🔹 Attributes Mapping:
 Vertex data → stress/temperature at a joint.

 Edge data → bond strength.

 Face data → ground cover in a map.

 Can encode data using color, opacity, texture, or glyphs (arrows, symbols, cylinders).

5.3.2 Visualizing Volume Data

Volume data = 3D equivalent of pixel data → represented as voxels (volume elements).
Sources:
 Acquired → CT scans, MRI, tomography.

 Simulated → Computational Fluid Dynamics (CFD).

Main Techniques:
1. Slicing (Cut Planes)

o Extract 2D slices from volume (row, column, depth).

o Can animate slices to see changes across layers.

o Supports non-planar slices, stacked slices, and orthogonal slices.

2. Isosurfaces (Marching Cubes Algorithm)

35
Dept. of CSE,SJBIT
23CSE532 Data Visualization Techniques

o Extracts surfaces corresponding to a constant value (isovalue).

o Works by checking voxel corners → inside/outside isovalue.

o 256 configurations possible, reduced to 14 unique by symmetry.

o Produces triangular patches forming continuous surfaces.

o Example: visualize density threshold in a 3D scalar field.

o ⚠️Challenges: large memory usage, voids in data, surface smoothness.

3. Direct Volume Rendering (DVR)

o No explicit polygons → computes pixel values directly.

o Two methods:

 Forward mapping (project voxels onto plane).

 Inverse mapping (Ray Casting) → send rays through volume and

accumulate voxel contributions.

o Requires resampling & interpolation.

o Compositing → accounts for opacity & transparency of voxels.

o Transfer Functions → map voxel values to color & opacity (highlight regions of
interest).

o Shading → estimated via gradients (central difference, Sobel operator) to

approximate surface normals.

5.3.3 Implicit Surfaces

Instead of explicitly defining surfaces, implicit surfaces define a function F(x,y,z) = 0.
 Advantages: good for blending, morphing, sparse data.

 Metaballs (Blobby Objects):

o Each particle has a density field decreasing with distance.

o Summing fields → smooth blending of shapes.

36
Dept. of CSE,SJBIT
23CSE532 Data Visualization Techniques

o Isosurface extraction creates fluid-like, organic shapes.

Examples:
 Blinn’s Gaussian Field: decays exponentially with distance.

 Wyvill’s Polynomial Field:Widely used in modeling molecules, soft bodies, and smooth
organic structures.

Summary Table
Method Representation Examples Pros Cons
Explicit Vertices, edges, faces, or Cube, Precise, easy to Needs consistent
Surfaces parametric equations cylinder manipulate orientation
Volume Data Shows internal Large data,
Sampling of 3D field MRI, CFD
(Voxels) structure interpolation needed
Medical Simple, May miss hidden
Slicing Cut planes
scans interactive features
Density Intuitive High memory &
Isosurfaces Constant-value surface
threshold surfaces complexity
Direct Volume Smoke, Rich detail, Computationally
Ray casting/compositing
Rendering fluids opacity effects heavy
Smooth
Implicit Harder to control
Zero-contour functions Metaballs blending,
Surfaces geometry
flexible

✅ This covers explicit surfaces, volume visualization (slicing, isosurface, DVR), and implicit
surfaces (metaballs) in detail.

🔹 Self-Learning Topics
 Munzner’s Nested Model:
o A modern framework for visualization design:
 Why? (Domain problem and task abstraction)
 What? (Data abstraction)

37
Dept. of CSE,SJBIT
23CSE532 Data Visualization Techniques

 How? (Encoding and interaction design)

 Where? (Algorithmic and system implementation)

🔹 Textbook Reference
 Chapter 4: Visualization Foundations, Visual Variables.
 Chapter 5: Visualization Techniques for Spatial Data.

🔹 Summary
Visualization depends on careful mapping of data to visual symbols.
 Eight visual variables guide how data is best represented.
 Taxonomies help select suitable techniques based on task and data type.
 1D, 2D, and 3D data each require appropriate visual methods.
 Tableau helps in creating visualizations and performing aggregations quickly.

Hands-on Exercise:
3. Basic Visualizations
● Create bar charts, line graphs, and tree maps.
●Explore the “Show Me” panel for quick visualizations.
● Implement basic filtering and sorting in views.

4. Aggregations & Calculations

● Use built-in aggregate functions (SUM, AVG, COUNT).
● Create calculated fields (e.g., profit margin, growth rate).
● Understand granularity and aggregation impacts.
3. Basic Visualizations
a. Bar Charts
 Drag Category → Columns.
 Drag Sales → Rows.
 Tableau automatically creates a bar chart.
38
Dept. of CSE,SJBIT
23CSE532 Data Visualization Techniques

 Switch between horizontal/vertical by swapping rows and columns.

b. Line Graphs
 Drag Order Date → Columns.
 Drag Sales → Rows.
 Change the Marks type to Line (or pick from Show Me).
 This gives a time-series sales trend.
c. Tree Maps
 Drag Category → Color/Label.
 Drag Sales → Size.
 Select Tree Map from Show Me.
 This shows the relative contribution of each category.
d. Show Me Panel
 Select fields (e.g., Region and Sales).
 Click different chart types from Show Me (bar, line, scatter, tree map, pie).
 This helps you experiment with visualizations quickly.
e. Filtering & Sorting
 Drag Region → Filters. Choose values to display.
 Right-click Sales axis → Sort Descending (highest sales first).
 Interactive filters can be shown using Show Filter (from field dropdown).

4. Aggregations & Calculations

a. Built-in Aggregate Functions
 Drag Sales → automatically aggregated as SUM(Sales).
 Right-click → change to AVG(Sales), COUNT(Orders), MIN, MAX.
 Example: Show average profit per category by dragging Profit → Rows and selecting
AVG.
b. Calculated Fields
1. Go to Analysis → Create Calculated Field.
2. Examples:
o Profit Margin = SUM([Profit]) / SUM([Sales])
o Growth Rate = (SUM([Sales]) - LOOKUP(SUM([Sales]), -1)) /
LOOKUP(SUM([Sales]), -1)
39
Dept. of CSE,SJBIT
23CSE532 Data Visualization Techniques

c. Granularity & Aggregation

 Granularity = Level of detail in data (e.g., Daily vs Monthly).
 Example: Sales by Day → highly detailed (spiky line).
 Aggregating to Month smooths the line.
 Use DATE_TRUNC or change date hierarchy to adjust.
 Remember: more granular = more rows, less aggregated = summarized.

40
Dept. of CSE,SJBIT