Visual Perception
Cecilia R. Aragon
I247
UC Berkeley
Spring 2010
Acknowledgments
• Thanks to slides and publications by Marti
Hearst, Pat Hanrahan, Christopher Healey,
Maneesh Agrawala, and Lawrence AndersonHuang, Colin Ware, Daniel Carr.
Spring 2010
I 247
2
Visual perception
•
•
•
•
•
•
•
•
Thinking with our Eyes
Structure of the Retina
Preattentive Processing
Detection
Estimating Magnitude
Change Blindness
Multiple Attributes
Gestalt
Spring 2010
I 247
3
Thinking with our Eyes
• 70% of body’s sense receptors reside in our eyes
• Metaphors to describe understanding often refer
to vision (“I see,” “insight,” “illumination”)
• “The eye and the visual cortex of the brain form a
massively parallel processor that provides the
highest-bandwidth channel into human cognitive
centers.” – Colin Ware, Information Visualization,
2004
• Important to understand how visual perception
works in order to effectively design visualizations
Spring 2010
I 247
4
Thinking with our Eyes
• Working memory is extremely limited
• How to overcome?
• “The processing of grouping simple concepts
into more complex ones is called chunking.” –
Ware, 2004
• “The process of becoming an expert is largely
one of learning to create effective chunks.” –
Ware, 2004
Spring 2010
I 247
5
The Power of Visualization
• “It is possible to have a far more complex
concept structure represented externally in a
visual display than can be held in visual and
verbal working memories.” – Ware, 2004
Spring 2010
I 247
6
How the Eye Works
The eye is not a camera!
Attention is selective (filtering)
Cognitive processes
Psychophysics: concerned with establishing
quantitative relations between physical
stimulation and perceptual events.
Spring 2010
I 247
7
How to Use Perceptual Properties
• Information visualization should cause what is meaningful to
stand out
Spring 2010
I 247
8
The Optimal Display
• Typical monitor: 35 pixels/cm
• = 40 cycles per degree at normal viewing
distances
• Human eye: receptors packed into fovea at
180 per degree of visual angle
• So a 4000x4000-pixel resolution monitor
should be adequate for most visual perception
tasks
Spring 2010
I 247
9
Optimal spatial resolution
• Humans can resolve a grating of 50 cycles per
degree (~44 pixels per cm)
• Sampling theory (Nyquist) states: need to
sample at twice the highest frequency needed
to detect
• So… why is 150 pixels per degree not sufficient
(cf. laser printers at 460 dots per cm)?
• 3 reasons: aliasing, gray levels, superacuities
• (will be discussed in future lecture)
Spring 2010
I 247
10
Structure of the Retina
Spring 2010
I 247
11
Structure of the Retina
• The retina is not a camera!
• Network of photo-receptor
cells (rods and cones) and
their connections
[Anderson-Huang, L.
http://www.physics.utoledo.edu/~lsa/
_color/18_retina.htm]
Spring 2010
I 247
12
Photo-transduction
• When a photon enters a receptor cell (e.g. a
rod or cone), it is absorbed by a molecule
called 11-cis-retinal and
converted to trans form.
• The different shape
causes it to ultimately
reduce the electrical
conductivity of the
photo-receptor cell.
[Anderson-Huang, L. http://www.physics.utoledo.edu/~lsa/_color/18_retina.htm]
Spring 2010
I 247
13
• Photoreceptors:
Retina
– 120 million rods, more sensitive than cones, not
sensitive to color
– 6-7 million cones, color sensitivity, concentrated in
macula (central 12 degrees of visual field)
– Fovea centralis - 2 degrees of visual field – twice
the width of thumbnail at arm’s length)
– Fovea comprises less
than 1% of retinal size
but 50% of visual cortex
Spring 2010
I 247
14
Electric currents from photo-receptors
• Photo-receptors generate an
electrical current in the dark.
• Light shuts off the current.
• Each doubling of light causes
roughly the same reduction of
current (3 picoAmps for cones, 6
for rods).
• Rods more sensitive, recover
more slowly.
• Cones recover faster, overshoot.
• Geometrical response in scaling
laws of perception.
[Anderson-Huang, L. http://www.physics
.utoledo.edu/~lsa/_color/18_retina.htm]
Spring 2010
I 247
15
Preattentive Processing
How many 5’s?
385720939823728196837293827
382912358383492730122894839
909020102032893759273091428
938309762965817431869241024
[Slide adapted from Joanna McGrenere http://www.cs.ubc.ca/~joanna/ ]
Spring 2010
I 247
17
How many 5’s?
385720939823728196837293827
382912358383492730122894839
909020102032893759273091428
938309762965817431869241024
Spring 2010
I 247
18
Preattentive Processing
• Certain basic visual properties are detected
immediately by low-level visual system
• “Pop-out” vs. serial search
• Tasks that can be performed in less than 200
to 250 milliseconds on a complex display
• Eye movements take at least 200 msec to
initiate
Spring 2010
I 247
19
Color (hue) is preattentive
• Detection of red circle in group of blue circles
is preattentive
[image from Healey 2005]
Spring 2010
I 247
20
Form (curvature) is preattentive
• Curved form “pops out” of display
[image from Healey 2005]
Spring 2010
I 247
21
Conjunction of attributes
• Conjunction target generally cannot be
detected preattentively (red circle in sea of
red square and blue circle distractors)
[image from Healey 2005]
Spring 2010
I 247
22
Healey
on preattentive processing
http://www.csc.ncsu.edu/faculty/healey/PP/index.html
Spring 2010
I 247
23
Preattentive Visual Features
line orientation
length
width
size
curvature
number
terminators
intersection
Spring 2010
closure
color (hue)
intensity
flicker
direction of motion
stereoscopic depth
3D depth cues
I 247
24
Cockpit dials
• Detection of a slanted line in a sea of vertical
lines is preattentive
Spring 2010
I 247
25
Detection
Spring 2010
I 247
26
Just-Noticeable Difference
• Which is brighter?
Spring 2010
I 247
27
Just-Noticeable Difference
• Which is brighter?
(130, 130, 130)
Spring 2010
(140, 140, 140)
I 247
28
Weber’s Law
• In the 1830’s, Weber made measurements of the
just-noticeable differences (JNDs) in the
perception of weight and other sensations.
• He found that for a range of stimuli, the ratio of
the JND ΔS to the initial stimulus S was relatively
constant:
ΔS / S = k
Spring 2010
I 247
29
Weber’s Law
• Ratios more important than magnitude in
stimulus detection
• For example: we detect the presence of a
change from 100 cm to 101 cm with the same
probability as we detect the presence of a
change from 1 to 1.01 cm, even though the
discrepancy is 1 cm in the first case and only
.01 cm in the second.
Spring 2010
I 247
30
Weber’s Law
• Most continuous variations in magnitude are
perceived as discrete steps
• Examples: contour maps, font sizes
Spring 2010
I 247
31
Estimating Magnitude
Spring 2010
I 247
32
Stevens’ Power Law
• Compare area of circles:
Spring 2010
I 247
33
Stevens’ Power Law
s(x) = axb
s is the sensation
x is the intensity of the
attribute
a is a multiplicative constant
b is the power
b > 1: overestimate
b < 1: underestimate
Spring 2010
[graph from Wilkinson 99]
I 247
34
Stevens’ Power Law
Sensation
Exponent
Brightness
0.33
Smell
0.55 (Coffee)
Loudness
0.6
Temperature
1.0 (Cold)
Taste
1.3 (Salt)
Heaviness
1.45
Electric Shock
3.5
[Stevens 1961]
Spring 2010
I 247
35
Stevens’ Power Law
Experimental results for b:
Length
Area
Volume
.9 to 1.1
.6 to .9
.5 to .8
Heuristic: b ~ 1/sqrt(dimensionality)
Spring 2010
I 247
36
Stevens’ Power Law
• Apparent magnitude scaling
[Cartography: Thematic Map Design, p. 170, Dent, 96]
S = 0.98A0.87
[J. J. Flannery, The relative effectiveness of some graduated point symbols in the presentation
of quantitative data, Canadian Geographer, 8(2), pp. 96-109, 1971] [slide from Pat Hanrahan]
Spring 2010
I 247
37
Relative Magnitude Estimation
Most accurate
Position (common) scale
Position (non-aligned) scale
Length
Slope
Angle
Least accurate
Spring 2010
Area
Volume
Color (hue/saturation/value)
I 247
38
Change Blindness
Spring 2010
I 247
39
Change Blindness
• An interruption in what is being seen causes
us to miss significant changes that occur in the
scene during the interruption.
• Demo from Ron Rensink:
http://www.psych.ubc.ca/~rensink/flicker/
Spring 2010
I 247
40
Possible Causes of Change Blindness
[Simons, D. J. (2000), Current approaches to change blindness,
Visual Cognition, 7, 1-16. ]
Spring 2010
I 247
41
Multiple Visual Attributes
Spring 2010
I 247
42
The Game of Set
•
•
•
•
Color
Symbol
Number
Shading
A set is 3 cards such that
each feature is EITHER
the same on each card
OR is different on each
card.
[Set applet by Adrien Treuille, http://www.cs.
washington.edu/homes/treuille/resc/set/]
Spring 2010
I 247
43
Multiple Visual Attributes
• Integral vs. separable
 Integral dimensions
 two or more attributes of an object are perceived holistically
(e.g.width and height of rectangle).
 Separable dimensions
 judged separately, or through analytic processing (e.g. diameter
and color of ball).
• Separable dimensions are orthogonal.
• For example, position is highly separable from
color. In contrast, red and green hue
perceptions tend to interfere with each other.
Spring 2010
I 247
44
Integral vs. Separable Dimensions
Integral
Separable
[Ware 2000]
Spring 2010
I 247
45
Gestalt
Spring 2010
I 247
46
Gestalt Principles
•
•
•
•
•
•
•
•
•
figure/ground
proximity
similarity
symmetry
connectedness
continuity
closure
common fate
transparency
Spring 2010
I 247
47
Examples
Proximity
Connectedness
[from Ware 2004]
Figure/Ground
[http://www.aber.ac.uk/media/Modules/MC10220/visper07.html]
Spring 2010
I 247
48
Conclusion
 What is currently known about visual
perception can aid the design process.
 Understanding low-level mechanisms of the
visual processing system and using that
knowledge can result in improved displays.
Spring 2010
I 247
49