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 ID: 933283
Download Presentation The PPT/PDF document "Visual Perception Cecilia R. Aragon" is the property of its rightful owner. Permission is granted to download and print the materials on this web site for personal, non-commercial use only, and to display it on your personal computer provided you do not modify the materials and that you retain all copyright notices contained in the materials. By downloading content from our website, you accept the terms of this agreement.
Slide1
Visual Perception
Cecilia R. Aragon
I247
UC Berkeley
Spring
2010
Slide2Acknowledgments
Thanks to slides and publications by
Marti Hearst, Pat
Hanrahan, Christopher Healey, Maneesh Agrawala, and Lawrence Anderson-Huang, Colin Ware, Daniel Carr.
Spring 2010
I 247
2
Slide3Visual perception
Thinking with our Eyes
Structure
of the RetinaPreattentive Processing Detection Estimating MagnitudeChange BlindnessMultiple Attributes
GestaltSpring
2010
I 247
3
Slide4Thinking 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, 2004Important to understand how visual perception works in order to effectively design visualizations
Spring 2010
I 247
4
Slide5Thinking 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
Slide6The 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
Slide7How the Eye Works
The eye is not a camera!
Attention is selective (filtering)
Cognitive processesPsychophysics: concerned with establishing quantitative relations between physical stimulation and perceptual events.
Spring
2010
I 247
7
Slide8How to Use Perceptual Properties
Information visualization should cause what is meaningful to stand out
Spring 2010
I 247
8
Slide9The 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 angleSo a 4000x4000-pixel resolution monitor should be adequate for most visual perception tasks
Spring 2010
I 247
9
Slide10Optimal 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 detectSo… 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
Slide11Structure of the Retina
Spring
2010
I 247
11
Slide12Structure of the Retina
The retina is
not
a camera!Network of photo-receptorcells (rods and cones) andtheir connections
Spring 2010
I 247
12
[Anderson-Huang, L. http://www.physics.utoledo.edu/~lsa/_color/18_retina.htm]
Slide13Photo-transduction
When a photon enters a receptor cell (e.g. a rod or cone), it is absorbed by a molecule called
11-cis-retinal
andconverted to trans form.The different shape
causes it to ultimatelyreduce the electricalconductivity of the
photo-receptor cell.Spring
2010
I 247
13
[Anderson-Huang, L. http://www.physics.utoledo.edu/~lsa/_color/18_retina.htm]
Slide14Retina
Photoreceptors:
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 lessthan 1% of retinal size
but 50% of visual cortex
Spring 2010
I 247
14
Slide15Electric 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.Spring
2010
I 247
15
[Anderson-Huang, L. http://www.physics
.
utoledo.edu
/~
lsa
/_color/18_retina.htm]
Slide16Preattentive Processing
Slide17How many 5’s?
385720939823728196837293827
382912358383492730122894839
909020102032893759273091428938309762965817431869241024
Spring 2010
I 247
17
[Slide adapted from Joanna
McGrenere
http://www.cs.ubc.ca/~joanna/ ]
Slide18How many 5’s?
38
5
720939823728196837293827382912358383492730122894839
909020102032893759273091428
938309762965817431869241024
Spring
2010
I 247
18
Slide19Preattentive 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 displayEye movements take at least 200 msec to initiateSpring
2010I 247
19
Slide20Color (hue) is preattentive
Detection of red circle in group of blue circles is preattentive
Spring
2010
I 247
20
[image from Healey 2005]
Slide21Form (curvature) is preattentive
Curved form “pops out” of display
Spring
2010
I 247
21
[image from Healey 2005]
Slide22Conjunction of attributes
Conjunction target generally cannot be detected
preattentively
(red circle in sea of red square and blue circle distractors)Spring 2010
I 247
22
[image from Healey 2005]
Slide23Healey
on
preattentive
processinghttp://www.csc.ncsu.edu/faculty/healey/PP/index.html
Spring 2010
I 247
23
Slide24Preattentive Visual Features
Spring
2010
I 247
24
line orientation
length
width
size
curvature
number
terminators
intersection
closure
color (hue)
intensity
flicker
direction of motion
stereoscopic depth
3D depth cues
Slide25Cockpit dials
Detection of a slanted line in a sea of vertical lines is preattentive
Spring
2010
I 247
25
Slide26Detection
Spring
2010
I 247
26
Slide27Just-Noticeable Difference
Which is brighter?
Spring
2010I 247
27
Slide28Just-Noticeable Difference
Which is brighter?
Spring
2010I 247
28
(130, 130, 130)
(140, 140, 140)
Slide29Weber’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
Slide30Weber’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
Slide31Weber’s Law
Most continuous variations in magnitude are perceived as discrete steps
Examples: contour maps, font sizes
Spring 2010
I 247
31
Slide32Estimating Magnitude
Spring
2010
I 247
32
Slide33Stevens’ Power Law
Compare area of circles:
Spring
2010I 247
33
Slide34Stevens’ Power Law
s(x) = ax
b
s is the sensationx is the intensity of the attribute
a is a multiplicative constantb is the powerb > 1: overestimate
b < 1: underestimate
Spring
2010
I 247
34
[graph from Wilkinson 99]
Slide35Stevens’ Power Law
Spring
2010
I 247
35
[Stevens 1961]
Slide36Stevens’ Power Law
Experimental results for
b
:Length .9 to 1.1Area .6 to .9Volume .5 to .8
Heuristic: b ~ 1/sqrt(dimensionality)
Spring 2010
I 247
36
Slide37Stevens’ Power Law
Apparent magnitude scaling
Spring
2010
I 247
37
[Cartography: Thematic Map Design, p. 170, Dent, 96]
S = 0.98A
0.87
[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
]
Slide38Relative Magnitude Estimation
Most accurate
Least accurate
Position (common) scale
Position (non-aligned) scale
Length
Slope
Angle
Area
Volume
Color
(hue/saturation/value)
Spring
2010
I 247
38
Slide39Change Blindness
Spring
2010
I 247
39
Slide40Change 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
Slide41Possible Causes of Change Blindness
Spring
2010
I 247
41
[Simons, D. J. (2000), Current approaches to change blindness, Visual Cognition, 7, 1-16. ]
Slide42Multiple Visual Attributes
Spring
2010
I 247
42
Slide43The Game of Set
Color
Symbol
NumberShading A set is 3 cards such that each feature is EITHER the same on each card OR is different on each card.
Spring 2010
I 247
43
[Set applet by
Adrien
Treuille
, http://www.cs.
washington.edu/homes/
treuille
/
resc
/set/]
Slide44Multiple 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
Slide45Integral vs. Separable Dimensions
Integral
Separable
Spring
2010
I 247
45
[Ware 2000]
Slide46Gestalt
Spring
2010
I 247
46
Slide47Gestalt Principles
figure/ground
proximity
similaritysymmetryconnectednesscontinuityclosurecommon fatetransparency
Spring 2010
I 247
47
Slide48Examples
Figure/Ground
Spring
2010
I 247
48
[http://www.aber.ac.uk/media/Modules/MC10220/visper07.html]
Proximity
Connectedness
[from Ware 2004]
Slide49Conclusion
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