Summary

[TOC]

week 2

basic concept

Imaging:
visual representation of an objects
Computer Graphics:
the process of creating pictures and films
Visualization:
the process of viewing, exploring, and transforming data as images
to gain understanding

A good visualization
abstract and explore the data effectively and efficiently

VR: visual representation

Data -> Visualization -> Human

Effective
easy to read, obtain; not confusing
Efficient
spend less time to count, understand

Human Visual Processing System

cornea 角膜 Light enters \
pupil 瞳孔 Control amount of light come through \
lens Focus on either near or distant objects \
virteous body 玻璃体 \
retina 视网膜 act as sensors – convert light to signals

photoreceptor cells (in retina)
Num: Rods > Cones

Rods: sensitive to low light condition
Cones: medium and bright light, color vision

retina blood vessels \
=> optic nerve(signal) \
=> visual cortex(images and visual perceptions) \
=> 3D formed with 2 eyes

eyes see in a high resolution only on a very small section \
of the field of view. \
This is because the high resolution is only in the fovea(中央凹).

inattentional blindness \
attend to task \
=> retain some portion of information (in brain)

Eye \
=> Low level Feature \
=> Pattern (higher features formed) \
=> Objects (limited info stored in memory)

Visual Queries

Acts of attention to find information to accomplish a task.

VIS guideline

Eyes beat memory
Make task and visual queries explicit
pre-attentive

week 3 && 4

Visual Channels

Visual Channel
Sender encodes information in signs \
Receiver decodes information from signs

Visual channels can be categorized to:

Expressiveness: what can be expressed via VC
Effectiveness: how well can be expressed via VC

Expressive

types of information

Quantitative information
position
Sequential information
position
Categorical information
color, shape

Effectiveness

Accuracy: precision of quantitative information.\
Discriminability: number of values can be
differentiated in a given visual channel.\
Salience: stand out from the rest.\
Separability: Interference btw. Channels.\
Grouping: elements are grouped formed a group.

Accuracy

Position > Length & Angle > Area

Position along a common scale
Position along unaligned scale (logarithm)
Length, direction, angle (1 D)
Area (2 D)
Volume, curvature (3 D + curve)
Shading, color saturation
Prioritize high-rank channels with reasons.

Discriminability

• Channel Properties (eg: line width)

• Spatial Arrangement (eg: aligned circle)\
• Size (eg: add axis)\
• Cardinality \
Low cardinality prefered
Strategies to encode high cardinality data

Grouping (less categories)
Filtering (top 5)
Faceting (seperate into diff plots)
1. Recreate the same type plots multiple times
2. Each plot represents only one specific category
3. All plots are aligned.

Salience

How well a VC can attract attention of viewers
Stand out from a scene
Preattentive Features
less than 200 to 250 milliseconds
• Preattentive Features \
color > size

line (blob) orientation
Size and length
Closure
Curvature
Intersection
3D depth cues
3D orientation
• Preattentive visual tasks \
Target detection:
detect “target” element
unique visual feature
Boundary detection:
detect boundary between two groups,
group common visual property
Region/object tracking:
track with a unique visual feature
• Non- Preattentive \
search serially
parallel line
Conjunctive Search (multiple channels)
Visual search for hidden objects

Separability

Interference btw. Channels for tuning attention

fully separable
position + hue
some interference
size + hue
significant interference
width + heigth + hue
width + height

Grouping

How easy a given channel
to represent information about groups \

similarity < proximity < connection & enclosure
similarity < continuity
• Proximity (closer together, eg: kNN) \
• Similarity (similar features, eg: same color or shape) \
• Connection (connected objects, eg: connected with wires) \
• Enclosure (enclosed by a contour, eg: bubble set) \
• Closure (missing information/part as whole, eg: IBM, Apple) \
• Continuity (follow a line or a curve, eg: function gragh) \

week 5

Color Perception & Color Spaces

Common use
1. Encode quantities with colors to detect pattern
2. Labelling – label data to distinguish categories
3. Highlight – draw attention
4. Color Brewer
5. Color scale (wave length B -> R)
6. Range of Values
  - Gray Scale
  - Intensity interpolation
  - Saturation interpolation
  - Two-color interpolation
  - Rainbow Scale
  - Heat Map interpolation
Color Perception
Photoreceptors cells
1. rods: night vision, 120m
2. cones: color vision, 6m
  L-cone: R 64%
  M: G 32%
  S: B 2~7%
  Blid spot: no photoreceptors
  optic nerve
  Color Measurement

Trichromatic Theory
(R,G,B).
Opponent Color Theory
process color as a three-channel system

Red-Green.
Blue-Yellow.
Black-White (luminance)
eg: stare R and then look at a neutral field, you see G
- use red, yellow and white to help attract attention
- not combine opponent

Color Specification

RGB space (hardware)
eg: colors change from green to blue \
more green, less blue
HSV(L) (human perception)
Hue: 0~360
Saturation: 0~100
Lightness/Value: 0~100
yellow: light, blue: dark
keep lightness then set saturation,
when L=100, any S => white

Is not perceptual uniform
CIE Lab(uv)
L brightness; a,b are chrominance components
a: + R, - G
b: + Y, - B
CIE Lch/ HCL
Hue: 0~360
Chroma: relative saturation 0~140
Lightness: 0~100
CMY and CMYK (color printing)
cyan (C), G + B
magenta (M), B + R
yellow (Y), R + G
key (K, black)

Color in Visualization

Quantitative Mapping (eg: Heat Map)
Quantitative Color Scale
- Uniformity
  value diff == perceptually color diff
  color used => not too light or dark(R, B attraction attention, need global perception)
- Discriminability
  many distinct values
Labelling
eg: categorize

Strategies

HCL space: human perception
fix chroma, lightness=> change other elements
sample uniformly

R, G, B, Y, White, Black easy to be distinguished

Gray:
no color, uncategorized, background, not important

Diverging Color Scales

red -> white -> green\
example: crop > medium height, < medium height\
some > sea level , < sea level\
some > abovr threshold, < below threshold

Expressiveness:

Quantative, Categorical
luminance changes on both sizes:
low luminance red -> white(high luminance) -> low luminancw green

week 9

Geographic Data Visualization

Geographic Data

Region
Geo-locations
Identifier(zipcode, street name)

spatial object

eg. Country, cities, lakes

Attributes:

Shape
Spatial locations

eg: states of USA, capital city of state in USA

object tracking && geo-localiztion

eg. track of car in video sequences

pixel-level georeferencing

from image pixel location => calculate the greography position \

when to use Map && why not use

Why

To represent a spatial phenomenon

Why Not

information not endcoded into map
quantity more inportant

use map
eg: cholera map
find the source of heisibing(cholera)

eg: too much infomation on map(many bank and city name)
the option: use matrix of 2 varaibles
city name\bank name bank1 bank2 ….
Landon
Paris
…

eg: area are small, hard to compare combination of whole regions
option: use stacked bar chart

Maps for visualiazation

Heat map
Dot map
Binned map
Choropleth map
Cartogram
Graduated symbol map

Heat map

depict continuous
density estimation method from discrete data
color to convey information (notice less rainbow map)

the method to estimate the kernel density function

$p(x) = \sum \frac{1}{\sqrt{2\pi}} e^{-(x- X_i)^2/(2\sigma^2)}$

KDE: kernel density estimation

Dot map

Dots represent geographical located specific objects
note: hard to see quantity (becuase each dot shape same)
easy to see density
Drawback:
Can’t see the amount of dots in the area

Hexbin Maps

fengchao(home of bees) 6 bianxing(hexagon)

dark: high quantity
light: low quantity

2D histogram map

use square instead of hexagon
Drawback:
hard to see clearly the shape
While hexgon easily representing shape, combination

Hexbin represent:

quantity
category (eg: map of UK)

Comparison of Square && Circle and Hexagon

circle could not fill the intervals among them; hexagon could
square:4 sides > hexagon: 6 sides:
flexibility of border:(90, 180, 270, 0 <=> 0, 60, 120, 180, 240…)
square has 8 neighbors: distance 1 or sqrt(2) \
hexagon has 6 neighbors: distance all the same 1
border is Not emphasized, while squares so

Choropleth map

shape are geographical region
regions are colored, patterned in relation of data variable

eg: map of France
types of data

data is spatial related
data is not raw (processed with stat)
unrestricted cover

points to notice using Cholo map

color uses
level of details
use nonlinear color map to express effectively

Categorical issue of Choro map:
drawback: quantative data is missing
can’t represent quantity in categorica based cholopleth map

note:
recommend diverging scale: eg: brewer \

perceptual uniform is important,

not choose colors randomly
not make boundary of 2 values: hard to see, huge difference

nonlinealy mapping Ok

contrast among colors
not all data => single hue sequential
=> diverging dynamic color

2 ways of mapping: choro map:

unclassed: mess up
classed: 0~10 the same color
Drawback: boudary is highlighted, when it is devided into big parts(it should be as background)

Cartogram

geometry of regions are distorted to convey info

dialation/shrink
erosion

for example, some properties of small region is extremely high => use circle area to reprensent properties => circle of it will overlap with others

Graduated Symbol map

size of symbol => represent quantity
not related to the size of regions

exmaple:
cities should not use it, dot map instead
so many cities, hard to see defference

eg. use pie chart(express percentages) as symbol

size
color
angle
temporal variable (1980~2000)
symbol (not shape)

Summary

Object:
- Dot map
Density:
- Heat map
- Binned map
Value:
- Choro map
- symbol map
- Cartogram map

Issue in maps

Base Rate Bias

more people => more library \
education level !=> more library

Sample size effect

Smaller size => higher variance
Bigger size => converge to the average of input data

skewed spatial distribution

eg: symbols are clustered in a small region (population shape in European countries)

Dorling Cartogram

shapes are repositioned

not maintain shape
Not maintain centroid

Perceptual Bias

eg: which area is highest value
When pallete is not designed well
(some objects share the “same” color for eyes)

eg: Context affect preception
background influence(contrast)
different background, same colored objects
=> seen different

different size of background objects, same size objects
=> seen different

interface in map features

use “few” visual channel to convey info via map
low color saturation
No strong lines/borders(or other info will be dismissed)

week 11

Visualizing Temporal Data

faceting
animation
slide: change a variable, the graph changes
line chart || bar chart || heat map

main temporal data

event data \
time + object
measurement data \
time + measure

line chart

improved cluttered lines: \
highlighting:

choose red to draw line
circle certain point with boundary

stack area charts:
show x1, x1+x2, x1+x2+x3, …

Aspect Ratio = Width/Height\
Average line slope should = 45’ degree

Heat map

Visualization for Event data

Gantt chart

timestamp + event properties

gantt
dateFormat  YYYY-MM-DD
title Adding GANTT diagram to mermaid
excludes weekdays 2014-01-10

section A section
Completed task            :done,    des1, 2014-01-06,2014-01-08
Active task               :active,  des2, 2014-01-09, 3d
Future task               :         des3, after des2, 5d
Future task2               :         des4, after des3, 5d

Another instance

gantt
    title A Gantt Diagram
    dateFormat  YYYY-MM-DD
    section Section
    A task           :a1, 2014-01-01, 30d
    Another task     :after a1  , 20d
    section Another
    Task in sec      :2014-01-12  , 12d
    another task      : 24d

Summary

week 2

basic concept

VR: visual representation

Human Visual Processing System

Visual Queries

VIS guideline

week 3 && 4

Visual Channels

Expressive

Effectiveness

Accuracy

Discriminability

Salience

Separability

Grouping

week 5

Color Perception & Color Spaces

Color Measurement

Color Specification

Color in Visualization

Strategies

Diverging Color Scales

week 9

Geographic Data Visualization

Geographic Data

spatial object

object tracking && geo-localiztion

pixel-level georeferencing

when to use Map && why not use

Maps for visualiazation

Heat map

Dot map

Hexbin Maps

2D histogram map

Choropleth map

Cartogram

Graduated Symbol map

Summary

Issue in maps

Base Rate Bias

Sample size effect

skewed spatial distribution

Dorling Cartogram

Perceptual Bias

interface in map features

week 11

Visualizing Temporal Data

main temporal data

category

line chart

Heat map

Visualization for Event data

Gantt chart